imperial college london · web viewincidence of dementia and major subtypes in europe: a...
TRANSCRIPT
Systematic evaluation of the associations between environmental risk factors and
dementia: an umbrella review of systematic reviews and meta-analyses
Vanesa Bellou1, Lazaros Belbasis1, Ioanna Tzoulaki1,2,3, Lefkos T Middleton4, John
PA Ioannidis5,6,7,8, Evangelos Evangelou1,2
1Department of Hygiene and Epidemiology, University of Ioannina Medical School,
Ioannina, Greece
2Department of Biostatistics and Epidemiology, School of Public Health, Imperial
College London, London, UK
3MRC-PHE Centre for Environment and Health, School of Public Health, Imperial
College London, London, UK
4Neuroepidemiology and Ageing Research Unit, School of Public Health, Imperial
College London, London, UK
5Department of Medicine, Stanford Prevention Research Center, Stanford, CA, USA
6Department of Health Research and Policy, Stanford University School of Medicine,
Stanford, CA, USA
7Meta-Research Innovation Center at Stanford (METRICS), Stanford University,
Stanford, CA, USA
8Department of Statistics, Stanford University School of Humanities and Sciences,
Stanford, CA, USA
1
Corresponding author:
Dr. Evangelos Evangelou, PhD
Department of Hygiene and Epidemiology, University of Ioannina Medical School,
Ioannina, 45110, Greece
e-mail: [email protected]
Role of the funding source
There was no funding source for this study. All authors had full access to all the study
data. The corresponding author had final responsibility for the decision to submit for
publication.
2
Abstract
INTRODUCTION: Dementia is a heterogeneous neurodegenerative disease, whose
aetiology results from a complex interplay between environmental and genetic
factors.
METHODS: We searched PubMed to identify meta-analyses of observational studies
that examined associations between non-genetic factors and dementia. We estimated
the summary effect size using random- and fixed-effects model, the 95% CI, and the
95% prediction interval. We assessed the between-study heterogeneity (I-square),
evidence of small-study effects, and excess significance.
RESULTS: 76 unique associations were examined. By applying standardised criteria,
7 associations presented convincing evidence. These associations pertained to
benzodiazepines use, depression at any age, late-life depression and frequency of
social contacts for all types of dementia; late-life depression for Alzheimer’s disease;
and type 2 diabetes mellitus for vascular dementia and Alzheimer’s diease.
DISCUSSION: Several risk factors present substantial evidence for association with
dementia and should be assessed as potential targets for interventions, but these
associations may not necessarily be causal.
Keywords: Alzheimer’s disease, dementia, epidemiology, risk factors, umbrella
review
Abbreviations: Alzheimer’s disease (AD), confidence interval (CI), interquartile
range (IQR), standard error (SE), type 2 diabetes mellitus (T2DM), prediction interval
(PI), vascular dementia (VaD)
3
Introduction
Over 46 million people live with dementia in 2016 world-wide and the number is
expected to exceed 130 million by 2050.1 This unprecedented increase of the number
of patients is mainly due to the considerable rise in life expectancy and population
ageing world-wide. The annual cost of dementia care was estimated at $818 billion,
world-wide, in 2014; and it is expected to exceed $1 trillion by 2018. No therapies are
currently available to delay or arrest the disease onset and progression and drug
development has been problematic, compared to other disease areas.2 Moreover, there
are considerable gaps in our understanding of the nosology, and aetiological
complexity of the disease.
Aimed at delaying disease onset by modulating modifiable risk factors, primary
prevention has been proposed as a potentially effective and feasible tool to address the
global challenge posed by dementia.3 It has been suggested that a third of Alzheimer’s
disease (AD) cases might be attributable to modifiable factors such as diabetes
mellitus, mid-life hypertension and obesity, physical activity, depression, smoking
and low educational attainment.4,5 An observed decline in the incidence and
prevalence of AD in western European countries and US has been ascribed to better
management of cardiovascular and metabolic risk factors.4,6–9 Unfortunately, it is
difficult to validate these speculations in randomized trials, since primary preventive
trials with clinical dementia outcomes would require large sample sizes and prolonged
follow-up. Due to the chronic and slowly progressive nature of this disease, both
pharmacological and non-pharmacological randomized clinical trials for dementia
mostly evaluate surrogate cognitive decline outcomes rather than clinical disease
outcomes.
4
We performed an umbrella review of the evidence across existing systematic reviews
and meta-analyses of observational studies to systematically map the evidence on
environmental risk factors for dementia. Our aim is to provide an overview of the
range and validity of the reported associations of diverse, potentially modifiable (non-
genetic), risk factors by evaluating whether there is evidence for biases in this
literature, and finally, pinpoint the number of previously studied associations that
have been synthesized with meta-analyses and have shown the strongest evidence for
association.
Methods
Search strategy and eligibility criteria
We conducted an umbrella review, i.e. a comprehensive and systematic collection and
evaluation of systematic reviews and meta-analyses performed on a specific research
topic.10 The methods of the umbrella review are standardized and follow the same
principles as previous umbrella reviews for other neurological disorders.11–13
We systematically searched PubMed up to January 16, 2016 to identify systematic
reviews and meta-analyses of observational studies examining associations of
potentially modifiable (environmental and other non-genetic) factors with all types of
dementia (Alzheimer’s disease, vascular dementia, dementia with Lewy bodies,
frontotemporal dementia). Relevant keywords for the search strategy were (dementia
OR Alzheimer*) AND (“systematic review” OR meta-analysis). Two independent
investigators (VB, LB) retrieved and abstracted the full text of potentially eligible
articles. We excluded meta-analyses that investigated the association between genetic
5
markers and risk for dementia as these factors have been examined extensively
elsewhere14,15 and they are not modifiable. We also did not consider fluid biomarkers
as they are not directly modifiable and the literature on fluid biomarkers is being
reviewed systematically elsewhere (http://www.alzforum.org/alzbiomarker). Meta-
analyses with an outcome related to cognitive decline or impairment, progression of
dementia or severity of symptoms were excluded. We further excluded meta-analyses
including less than three component studies. When an association was covered by
more than one meta-analyses, we kept the meta-analysis with the largest number of
component studies with available data on individual studies. We did not apply any
language restrictions in our search strategy.
Data extraction
Two independent investigators (VB, LB) extracted the data, and in case of
discrepancies consensus was reached. From each eligible article, we abstracted
information on the first author, journal and year of publication, the examined risk
factors and the number of studies considered. We also extracted the study-specific
risk estimates (i.e. risk ratio, odds ratio, hazard ratio) along with their corresponding
confidence interval (CI) and the number of cases and controls in each study. If a risk
factor was examined in more than one levels of comparison, we extracted the data for
the comparison having the largest number of component studies. Also, when a meta-
analysis combined effect estimates for incidence of dementia and score in a cognitive
test, we considered the former. Furthermore, we recorded whether the eligible papers
applied any criteria to assess the quality of component studies.
Statistical analysis
6
We applied standardized methods for the umbrella review and state-of-the-art
approaches to evaluate findings on putative risk factors for dementia, that have been
applied to assess the epidemiological credibility for environmental risk factors of
other neurodegenerative diseases,11–13 while similar assessments have been
successfully applied in genetic studies.16,17 Specifically, for each meta-analysis, we
estimated the summary effect size and its 95% CI using both fixed-effects and
random-effects models.18,19 We also estimated the 95% prediction interval (PI), which
accounts for the between-study heterogeneity and evaluates the uncertainty for the
effect that would be expected in a new study addressing that same association.20,21 For
the largest study of each meta-analysis, we estimated the standard error (SE) of the
effect size and we examined whether the SE was less than 0.10. In a study with SE of
less than 0.10, the difference between the effect estimate and the upper or lower 95%
CI is less than 0.20 (i.e. this uncertainty is less than what is considered a small effect
size).
Between-study heterogeneity was quantified using the I2 metric.22 I2 ranges between
0% and 100% and quantifies the variability in effect estimates that is due to
heterogeneity rather than sampling error.23 Values exceeding 50% or 75% are
considered to represent large or very large heterogeneity, respectively.
We assessed small-study effects (i.e. whether smaller studies tend to give
substantially larger estimates of effect size compared with larger studies)24 using the
Egger’s regression asymmetry test.25 A p<0.10 combined with a more conservative
effect in the largest study than in random-effects meta-analysis was judged to provide
adequate evidence for small-study effects.
7
We further applied the excess statistical significance test, which evaluates whether
there is a relative excess of formally significant findings in the published literature
due to any reason (e.g. publication bias, selective reporting of outcomes or analyses).
It is a chi-square based test that assesses whether the observed (O) number of studies
with nominally significant results is larger than their expected (E) number.26 We used
the effect size of the largest study (smallest SE) in each meta-analysis to calculate the
power of each study using a non-central t distribution.27,28 Excess statistical
significance was claimed at two-sided p<0.10 with O>E as previously proposed.26
Assessment of epidemiological credibility
We identified associations that had the strongest evidence and no signals of large
heterogeneity or bias. Specifically, we considered as convincing the associations that
fulfilled all of the following criteria: statistical significance according to random-
effects model at p<10-6;29,30 based on more than 1,000 cases; without large between-
study heterogeneity (I2<50%); 95% PI excluding the null value; and no evidence of
small-study effects and excess significance. Associations with >1000 cases, p<10-6
and largest study presenting a statistically significant effect were graded as highly
suggestive. The associations supported by >1,000 cases and a significant effect at
p<10-3 were considered as suggestive. The remaining nominally significant
associations were considered as having weak evidence.
Association does not necessarily imply causation. For associations with convincing or
highly suggestive evidence, we performed a sensitivity analysis including only
prospective cohort studies in order to assess whether there is evidence also for
temporality of the association. We also qualitatively discuss the evidence on these
associations in terms of the potential for reverse causation, residual confounding,
8
information bias (e.g. non-differential misclassification), and other biases not covered
by the standardized tests listed above.
The statistical analysis and the power calculations were done with STATA version
12.0.
Results
Overall, 2543 papers were searched and 43 articles were eligible. Of the 120 papers
screened in full text, 40 were excluded because a larger meta-analysis on the same
association was found, 32 because they had no quantitative synthesis and 5 for other
reasons (Figure 1). A list of the 40 excluded meta-analyses is presented in the
Supplementary Table 1.
The eligible papers were published between 2008 and 2016. The 43 articles
corresponded to 76 unique meta-analyses: 39 on Alzheimer’s disease (AD), 27 on all
types of dementia, and 10 on vascular dementia (VaD). Fifty-three unique risk factors
were considered and 14 of them were studied in more than one outcome. The median
number of studies per meta-analysis was 7 (IQR, 5-13) and the median number of
cases was 1,139 (IQR, 590-3,537). The number of cases was greater than 1,000 in 46
meta-analyses (Table 1). All eligible meta-analyses used summary-level data from
published literature and none of them had access to individual participant data.
Ten of the eligible articles used the Newcastle-Ottawa scale to qualitatively assess
259 component studies. Of the 259 studies assessed, 63 (24%) were of low quality,
124 (48%) were graded as having moderate quality, and 72 (28%) were characterized
as high quality (Supplementary Table 2).
9
Associations for AD
39 of the 76 meta-analyses examined associations for AD. 26 of 39 meta-analyses
(67%) presented a nominally statictically significant effect (p<0.05), and only 6
associations (15%) reached p<10-6 (Table 1). These associations pertained to cancer,
depression at any age, late-life depression, neuroticism, physical activity, and type 2
diabetes mellitus (T2DΜ). 7 of 39 associations (aluminum, cancer, late-life
depression, neuroticism, physical activity, stroke, T2DΜ) had a 95% PI excluding the
null value (Table 1). 15 associations presented large between-study heterogeneity
estimates (I2≥50% and I2≤75%), and three associations presented very large
heterogeneity estimates (I2>75%) (Table 1). 8 associations presented evidence for
small-study effects and 12 associations had a statistically significant excess of
“positive” studies (Table 1, Supplementary Table 3).
Associations for all types of dementia
27 of the 76 meta-analyses explored associations for all types of dementia. 20 of the
27 associations (74%) had a statictically significant effect at p<0.05, and only 6 of
them were statistically significant when we applied a more stringent threshold (p<10-
6) (Table 1). These associations pertained to benzodiazepines use, depression at any
age, education, frequency of social contacts, late-life depression, and T2DΜ. In 6
associations (depression at any age, early-life depression, late-life depression, T2DΜ,
low frequency of social contacts and benzodiazepines use), the 95% PI under random-
effects model excluded the null value (Table 1). 15 associations presented large or
very large heterogeneity estimates (I2≥50%) (Table 1). Of the 27 associations for all
types of dementia, 5 had evidence for small-study effects (antihypertensive drugs, fish
intake, physical activity, social participation, and statins) and 7 had a statistically
10
significant excess of “positive” studies (alcohol drinking, education, physical activity,
rural living, smoking, social participation, and tooth loss) (Table 1, Supplementary
Table 3).
Associations for VaD
Only 10 of the 76 associations examined a risk factor for VaD. Two associations
(smoking and T2DM) were supported by more than 1,000 cases. Nine associations
had a nominally statistically significant effect (p<0.05), while three of them
(education, late-life depression, and T2DM) presented a p<10-6 (Table 1). Three
associations (depression at any age, late-life depression and T2DM) had a 95% PI
excluding the null value (Table 1). One association (physical activity) had large
heterogeneity, and another association (midlife obesity) had very large heterogeneity
(Table 1). One association (smoking) had evidence for excess significance, and none
of the associations had evidence for small-study effects (Table 1, Supplementary
Table 3).
Assessment of epidemiological credibility
Βy applying the predefined methodological criteria, four risk factors for all types of
dementia (benzodiazepines use, depression at any age, late-life depression and
frequency of social contacts), two risk factors for AD (late-life depression, T2DM)
and one risk factor for VaD (T2DM) were supported by convincing evidence, by
having more than 1,000 cases, a p<10-6, not large heterogeneity (I2<50%), 95% PI
excluding the null value, no hints for small-study effects and excess significance bias.
Moreover, the association of T2DM with all types of dementia, and the association of
cancer, depression at any age and physical activity with AD were supported by highly
suggestive evidence (more than 1,000 cases, p<10-6 and a nominally significant effect
11
in the largest study). Furthermore, five risk factors for AD (aluminum, herpesviridae
infection, low-frequency electromagnetic fields, education, NSAIDs) and five risk
factors for all types of dementia (early-life depression, education, midlife obesity,
physical activity, statins) presented suggestive evidence. Additional 34 associations
presented weak evidence with p<0.05. Finally, 21 associations did not present even a
nominally statistically significant result (p>0.05) (Table 2).
In a sensitivity analysis, limited to prospective cohort studies only, the summary
effect size and the p-value remained similar for all risk factors with convincing and
highly suggestive evidence (Table 3).
Discussion
We systematically collected and appraised the potentially modifiable environmental
risk factors that have previously been reported for association with dementia in
observational studies and summarized in meta-analyses. Only depression at any age,
late-life depression, benzodiazepines use, and low frequency of social contacts
showed convincing epidemiological evidence for an increased risk for all types of
dementia. For AD only late-life depression and T2DM showed convincing evidence.
T2DM was the only risk factor with convincing evidence for an association with VaD,
while the association of T2DM with all types of dementia was supported by highly
suggestive evidence. Also, the association of history of cancer, depression at any age
and physical activity with risk for AD was supported by highly suggestive evidence.
Analyses limited to prospective cohort studies gave very similar estimates for these
risk factors.
12
Half of the examined meta-analyses had large heterogeneity and one third suffered
from small-study effects or/and excess significance. Heterogeneity could be due to
biased results in some of the included studies, but it could also reflect genuine
differences across studies. Sources of heterogeneity in dementia research could be
introduced from different study designs (i.e. a mixture of prospective and
retrospective studies in the meta-analyses) and differences in exposure assessment.
Reported associations with disease should be interpreted with caution, in particular
when heterogeneity is large and small-study effects are evident.
Furthermore, the diagnostic criteria and definition of AD and, indeed, late-onset
dementia remains a controvertial issue. There is still no consensus in diagnostic
criteria used for dementia diagnosis. This could create between-study heterogeneity
and hinder the identification of robust risk factors. Indeed, the National Institute of
Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease
and Related Disorders Association 1984 diagnostic criteria for AD employed the
terms of clinically probable and possible AD, with the assumption that a definite
diagnosis requires post mortem confirmation of the typical AD-like pathology.31 The
recently revised criteria by the National Institute of Aging and Alzheimer’s
Association workgroup have differentiated all-cause dementia from AD, with (as
exclusion criterion) the presence of any clinical or other evidence of concomitant
cerebro-vascular pathology, any other known form of dementia or any concurrent
condition or medication potentially affecting cognition.32 However, whilst 70% of
dementia cases occur in people over the age of 75,33 the majority of AD patients over
the age of 75 have co-existing pathologies on brain biopsy.34
Depression at any age and late-life depression showed convincing evidence of
association with all types of dementia. It is disputed whether depression is a risk
13
factor or a mere prodrome of dementia.35 To test this hypothesis, an analysis taking
into account the relative association between age at onset of depression and age at
diagnosis of dementia could more precisely define the association. Indeed, late-life
depression presented the strongest and most consistent evidence supporting an
increased risk for both all types of dementia and AD.36 On the other hand, early-life
depression yielded a significant association supported by suggestive evidence, while
only one non-significant prospective cohort study examined the association between
early-life depression and all types of dementia.35 It has also been postulated that
depression may be an early reaction to perceived cognitive decline. Our analysis also
suggests a higher risk for VaD in patients with late-life depression but the supporting
evidence was sparse, partly due to the smaller number of studies in the meta-analysis
of VaD compared to AD. According to the “vascular depression hypothesis”, vascular
disease, leading to a long-term process of subclinical cerebrovascular changes, could
be the underlying link between depression and dementia.37
Additionally, our umbrella review reveals a strong relationship of T2DΜ with VaD
and AD. VaD has been classified, based on the pattern of cerebrovascular lesions, into
multi-infarct dementia, strategic infarct dementia and subcortical vascular
encephalopathy (Binswanger’s disease).38,39 These disease patterns indicate that VaD
and cerebrovascular diseases might share common risk factors, including T2DΜ.40
The highly suggestive association of T2DΜ with all types of dementia presented large
between-study heterogeneity. This may reflect that T2DΜ confers a different
magnitude of susceptibility for different types of dementia, with more than doubling
of the risk for VaD, but a more modest increase in the risk for AD.
High level of physical activity has a protective effect for the risk of developing AD in
prospective cohort studies. No heterogeneity was observed and the 95% PI excluded
14
the null value. However, the summary effect size and its significance level may be
affected by the presence of small-study effects. In support, two RCTs have suggested
that physical activity interventions can potentially reduce the cognitive decline in
elderly people.41,42
Furthermore, a statistically significant association supported by highly suggestive
evidence was observed for history of cancer and AD. Although the summary effect
estimate remained similar in the sensitivity analysis of the prospective cohort studies,
hints for small-study effects were present. Also, given that the patients with cancer
tend to have shorter life expectancy than the general population, this association could
be attributed to the competing risks between death and risk for AD.43 Finally, the
component studies used adjustment models mainly for age and sex and they did not
perform adjustment for potential confounders, like history of depression and T2DΜ.
These are significant limitations indicating that the reported association between
cancer and AD might not be genuine.
Moreover, low frequency of social contacts showed a statistically significant effect on
developing all types of dementia. Social networking is thought to be among the
modifiable factors, along with the level of educational attainment, and a range of
leisure activities, which may help maintain cognitive function in old age. This concept
of “brain reserve” refers to the ability to tolerate the age-related changes and the
disease-related pathology in the brain without developing clear clinical symptoms or
signs.44, The level of education, another factor associated with the brain reserve
theory, had suggestive evidence for an association with all types of dementia and
weak evidence for an association with AD. Both associations were highly significant,
but they presented very large between-study heterogeneity and wide 95% PI including
the null value. The observational studies might apply different diagnostic criteria and
15
different methods to measure educational attainment. These factors are potential
sources of the observed between-study heterogeneity, along with other differences in
study design (e.g., variability of adjustment models). Also, the association between
level of education and VaD was based on a small number of cases, and thus this
association was not considered to be supported by strong evidence.
The increased risk of developing dementia for those who use benzodiazepines could
be the result of limited cognitive reserve capacity, induced by the long term use of
benzodiazepines, which might reduce a person’s ability to cope with early phase brain
lesions by soliciting accessory neuronal networks.45 Of note, the cohort studies that
evaluated the association between benzodiazepines use and dementia had a
prospective design, which argues against reverse causation. However,
benzodiazepines use might have been motivated by the presence of symptoms, such
as difficulties with sleep and chronic anxiety with or without depression, which are
known psychopathological features that may precede the formal diagnosis of
dementia by years.46 Additionally, a recently published cohort study indicated the
absence of a dose-response association between use of benzodiazepines and risk for
dementia.47 This observation argues against a causative association.
Our study has several limitations. The diagnostic criteria used for case ascertainment
could influence the effect size and increase between-study heterogeneity. We could
not fully control for these factors because this information was, often, not available in
the published meta-analyses. Also, it is apparent that the mixture of both low and
high-quality studies influences the summary effect size in a meta-analysis; thus, in
meta-analyses the quantitative synthesis of the evidence should follow a critical
appraisal of the pertinent observational studies. We could not perform such an
appraisal, as this process should have been a main objective of the original meta-
16
analyses. Indeed, only a quarter of the eligible papers assessed and reported the study
quality using the Newcastle-Ottawa scale, a standardized tool for qualitative
assessment of non-randomized studies. Based on this qualitative appraisal, only a
quarter of the component studies presented high methodological quality and were of
low risk for bias. Finally, in our analysis we considered only associations that have
been evaluated in meta-analyses of observational studies and we might have missed
associations with adequate evidence that have not yet been assessed through meta-
analytic quantitative synthesis.
Taking into account these caveats, our analysis identified five risk factors with
convincing evidence and strong epidemiological credibility pertaining to
benzodiazepines use, depression at any age, low frequency of social contacts, T2DΜ
and late-life depression. The ability to modify these factors through public health
policy measures or other interventions remains to be established. Indeed, there is no
guarantee that even a convincing association in observational evidence for a
modifiable risk factor would necessarily translate into large preventive benefits for
dementia if an effort is made to modify these risk factors.48 Currently there are means
available that can achieve substantial improvements in the control of T2DM, but the
burden of T2DM is increasing in many developed and developing countries. Policy
interventions should also be feasible aiming at limitation of benzodiazepines use in
the older population. The ability to reduce the population burden of depression or
enhance social interactions is more limited. The continued search for additional risk
factors for dementia and their validation in real life are still warranted.
17
References
1 Alzheimer’s Disease International. World Alzheimer’s report 2015: The global impact of dementia. London, 2015.
2 Gauthier S, Albert M, Fox N, et al. Why has therapy development for dementia failed in the last two decades? Alzheimer’s Dement 2016; 12: 60–4.
3 G8 Health and Science Ministers. G8 Dementia Summit Declaration. London, 2013.
4 Norton S, Matthews FE, Barnes DE, Yaffe K, Brayne C. Potential for primary prevention of Alzheimer’s disease: An analysis of population-based data. Lancet Neurol 2014; 13: 788–94.
5 Barnes DE, Yaffe K. The projected effect of risk factor reduction on Alzheimer’s disease prevalence. Lancet Neurol 2011; 10: 819–28.
6 Qiu C, von Strauss E, Bäckman L, Winblad B, Fratiglioni L. Twenty-year changes in dementia occurrence suggest decreasing incidence in central Stockholm, Sweden. Neurology 2013; 80: 1888–94.
7 Schrijvers EMC, Verhaaren BFJ, Koudstaal PJ, Hofman A, Ikram MA, Breteler MMB. Is dementia incidence declining?: Trends in dementia incidence since 1990 in the Rotterdam Study. Neurology 2012; 78: 1456–63.
8 Matthews FE, Arthur A, Barnes LE, et al. A two-decade comparison of prevalence of dementia in individuals aged 65 years and older from three geographical areas of England: results of the Cognitive Function and Ageing Study I and II. Lancet (London, England) 2013; 382: 1405–12.
9 Satizabal CL, Beiser AS, Chouraki V, Chêne G, Dufouil C, Seshadri S. Incidence of Dementia over Three Decades in the Framingham Heart Study. N Engl J Med 2016; 374: 523–32.
10 Ioannidis JPA. Integration of evidence from multiple meta-analyses: a primer on umbrella reviews, treatment networks and multiple treatments meta-analyses. CMAJ 2009; 181: 488–93.
11 Belbasis L, Bellou V, Evangelou E, Ioannidis JPA, Tzoulaki I. Environmental
18
risk factors and multiple sclerosis: an umbrella review of systematic reviews and meta-analyses. Lancet Neurol 2015; 14: 263–73.
12 Bellou V, Belbasis L, Tzoulaki I, Evangelou E, Ioannidis JPA. Environmental risk factors and Parkinson’s disease: An umbrella review of meta-analyses. Parkinsonism Relat Disord 2016; 23: 1–9.
13 Belbasis L, Bellou V, Evangelou E. Environmental Risk Factors and Amyotrophic Lateral Sclerosis: An Umbrella Review and Critical Assessment of Current Evidence from Systematic Reviews and Meta-Analyses of Observational Studies. Neuroepidemiology 2016; 46: 96–105.
14 Bertram L, McQueen MB, Mullin K, Blacker D, Tanzi RE. Systematic meta-analyses of Alzheimer disease genetic association studies: the AlzGene database. Nat Genet 2007; 39: 17–23.
15 Dwyer R, Skrobot OA, Dwyer J, Munafo M, Kehoe PG. Using Alzgene-like approaches to investigate susceptibility genes for vascular cognitive impairment. J Alzheimer’s Dis 2013; 34: 145–54.
16 Ioannidis JPA, Boffetta P, Little J, et al. Assessment of cumulative evidence on genetic associations: Interim guidelines. Int J Epidemiol 2008; 37: 120–32.
17 Belbasis L, Panagiotou OA, Dosis V, Evangelou E. A systematic appraisal of field synopses in genetic epidemiology: a HuGE review. Am J Epidemiol 2015; 181: 1–16.
18 DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7: 177–88.
19 Lau J, Ioannidis JP, Schmid CH. Quantitative synthesis in systematic reviews. Ann Intern Med 1997; 127: 820–6.
20 Higgins JPT, Thompson SG, Spiegelhalter DJ. A re-evaluation of random-effects meta-analysis. J R Stat Soc Ser A Stat Soc 2009; 172: 137–59.
21 Higgins JPT. Commentary: Heterogeneity in meta-analysis should be expected and appropriately quantified. Int J Epidemiol 2008; 37: 1158–60.
22 Cochran WG. The Combination of Estimates from Different Experiments. Biometrics 1954; 10: 101.
19
23 Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med 2002; 21: 1539–58.
24 Sterne JAC, Sutton AJ, Ioannidis JPA, et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ 2011; 343: d4002.
25 Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315: 629–34.
26 Ioannidis JPA, Trikalinos TA. An exploratory test for an excess of significant findings. Clin Trials 2007; 4: 245–53.
27 Ioannidis JPA. Clarifications on the application and interpretation of the test for excess significance and its extensions. J Math Psychol 2013; 57: 184–7.
28 Lubin JH, Gail MH. On power and sample size for studying features of the relative odds of disease. Am J Epidemiol 1990; 131: 552–66.
29 Ioannidis JPA, Tarone R, McLaughlin JK. The false-positive to false-negative ratio in epidemiologic studies. Epidemiology 2011; 22: 450–6.
30 Johnson VE. Revised standards for statistical evidence. Proc Natl Acad Sci U S A 2013; 110: 19313–7.
31 McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology 1984; 34: 939–44.
32 McKhann GM, Knopman DS, Chertkow H, et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 263–9.
33 Fratiglioni L, Launer LJ, Andersen K, et al. Incidence of dementia and major subtypes in Europe: A collaborative study of population-based cohorts. Neurologic Diseases in the Elderly Research Group. Neurology 2000; 54: S10–5.
34 Schneider JA, Arvanitakis Z, Bang W, Bennett DA. Mixed brain pathologies account for most dementia cases in community-dwelling older persons.
20
Neurology 2007; 69: 2197–204.
35 Byers AL, Yaffe K. Depression and risk of developing dementia. Nat Rev Neurol 2011; 7: 323–31.
36 Diniz BS, Butters M a, Albert SM, Dew MA, Reynolds CF. 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 2013; 202: 329–35.
37 Barnes DE, Yaffe K, Byers AL, McCormick M, Schaefer C, Whitmer RA. Midlife vs Late-Life Depressive Symptoms and Risk of Dementia: Differential Effects for Alzheimer Disease and Vascular Dementia. Arch Gen Psychiatry 2012; 69: 493–8.
38 O’Brien JT, Thomas A. Vascular dementia. Lancet 2015; 386: 1698–706.
39 Jellinger KA. Morphologic diagnosis of ‘vascular dementia’ - a critical update. J Neurol Sci 2008; 270: 1–12.
40 Gorelick PB, Scuteri A, Black SE, et al. Vascular Contributions to Cognitive Impairment and Dementia: A Statement for Healthcare Professionals From the American Heart Association/American Stroke Association. Stroke 2011; 42: 2672–713.
41 Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. Lancet (London, England) 2015; 385: 2255–63.
42 Lautenschlager NT, Cox KL, Flicker L, Foster JK, Bockxmeer FM Van. Effect of Physical Activity on Cognitive Function in Older Adults at Risk for Alzheimer Disease. JAMA 2008; 300: 1027–37.
43 Zhang Q, Guo S, Zhang X, et al. Inverse relationship between cancer and Alzheimer’s disease: a systemic review meta-analysis. Neurol Sci 2015; 36: 1987–94.
44 Fratiglioni L, Wang H-X. Brain reserve hypothesis in dementia. J Alzheimers Dis 2007; 12: 11–22.
45 Billioti de Gage S, Moride Y, Ducruet T, et al. Benzodiazepine use and risk of
21
Alzheimer’s disease: case-control study. BMJ 2014; 349: g5205.
46 Barbui C, Gastaldon C, Cipriani A. Benzodiazepines and risk of dementia: true association or reverse causation? Epidemiol Psychiatr Sci 2013; 22: 307–8.
47 Gray SL, Dublin S, Yu O, et al. Benzodiazepine use and risk of incident dementia or cognitive decline: prospective population based study. BMJ 2016; 352: i90.
48 Prasad V, Jorgenson J, Ioannidis JPA, Cifu A. Observational studies often make clinical practice recommendations: an empirical evaluation of authors’ attitudes. J Clin Epidemiol 2013; 66: 361–6.e4.
49 Anstey KJ, Mack HA, Cherbuin N. Alcohol consumption as a risk factor for dementia and cognitive decline: meta-analysis of prospective studies. Am J Geriatr Psychiatry 2009; 17: 542–55.
50 Anstey KJ, Cherbuin N, Budge M, Young J. Body mass index in midlife and late-life as a risk factor for dementia: a meta-analysis of prospective studies. Obes Rev 2011; 12: e426–37.
51 Beckett MW, Ardern CI, Rotondi MA. A meta-analysis of prospective studies on the role of physical activity and the prevention of Alzheimer’s disease in older adults. BMC Geriatr 2015; 15: 9.
52 Cataldo JK, Prochaska JJ, Glantz SA. Cigarette smoking is a risk factor for Alzheimer’s Disease: an analysis controlling for tobacco industry affiliation. J Alzheimers Dis 2010; 19: 465–80.
53 Chang-Quan H, Hui W, Chao-Min W, et al. The association of antihypertensive medication use with risk of cognitive decline and dementia: a meta-analysis of longitudinal studies. Int J Clin Pract 2011; 65: 1295–305.
54 da Silva J, Goncalves-Pereira M, Xavier M, Mukaetova-Ladinska EB. Affective disorders and risk of developing dementia: systematic review. Br J Psychiatry 2013; 202: 177–86.
55 García AM, Sisternas A, Hoyos SP. Occupational exposure to extremely low frequency electric and magnetic fields and Alzheimer disease: A meta-analysis. Int J Epidemiol 2008; 37: 329–40.
56 Gudala K, Bansal D, Schifano F, Bhansali A. Diabetes mellitus and risk of
22
dementia: A meta-analysis of prospective observational studies. J Diabetes Investig 2013; 4: 640–50.
57 Kim Y-S, Kwak SM, Myung S-K. Caffeine intake from coffee or tea and cognitive disorders: a meta-analysis of observational studies. Neuroepidemiology 2015; 44: 51–63.
58 Li F-J, Shen L, Ji H-F. Dietary intakes of vitamin E, vitamin C, and β-carotene and risk of Alzheimer’s disease: a meta-analysis. J Alzheimers Dis 2012; 31: 253–8.
59 Maheshwari P, Eslick GD. Bacterial infection and Alzheimer’s disease: a meta-analysis. J Alzheimers Dis 2015; 43: 957–66.
60 Meng X, D’Arcy C. Education and Dementia in the Context of the Cognitive Reserve Hypothesis: A Systematic Review with Meta-Analyses and Qualitative Analyses. PLoS One 2012; 7: e38268.
61 Meng X-F, Yu J-T, Wang H-F, et al. Midlife vascular risk factors and the risk of Alzheimer’s disease: a systematic review and meta-analysis. J Alzheimers Dis 2014; 42: 1295–310.
62 O’Brien J, Jackson JW, Grodstein F, Blacker D, Weuve J. Postmenopausal Hormone Therapy Is Not Associated With Risk of All-Cause Dementia and Alzheimer’s Disease. Epidemiol Rev 2014; 36: 83–103.
63 Perry DC, Sturm VE, Peterson MJ, et al. Association of traumatic brain injury with subsequent neurological and psychiatric disease: a meta-analysis. J Neurosurg 2015; : 1–16.
64 Peters R, Booth A, Peters J. A systematic review of calcium channel blocker use and cognitive decline/dementia in the elderly. J Hypertens 2014; 32: 1945–57; discussion 1957–8.
65 Power MC, Weuve J, Gagne JJ, McQueen MB, Viswanathan A, Blacker D. The association between blood pressure and incident Alzheimer disease: a systematic review and meta-analysis. Epidemiology 2011; 22: 646–59.
66 Richardson K, Schoen M, French B, et al. Statins and cognitive function: a systematic review. Ann Intern Med 2013; 159: 688–97.
67 Seitz DP, Shah PS, Herrmann N, Beyene J, Siddiqui N. Exposure to general
23
anesthesia and risk of alzheimer’s disease: a systematic review and meta-analysis. BMC Geriatr 2011; 11: 83.
68 Steel AJ, Eslick GD. Herpes Viruses Increase the Risk of Alzheimer’s Disease: A Meta-Analysis. J Alzheimers Dis 2015; 47: 351–64.
69 Terracciano A, Sutin AR, An Y, et al. Personality and risk of Alzheimer’s disease: New data and meta-analysis. Alzheimer’s Dement 2014; 10: 179–86.
70 Virk SA, Eslick GD. Brief Report: Meta-analysis of Antacid Use and Alzheimer’s Disease: Implications for the Aluminum Hypothesis. Epidemiology 2015; 26: 769–73.
71 Wang J, Tan L, Wang H-F, et al. Anti-inflammatory drugs and risk of Alzheimer’s disease: an updated systematic review and meta-analysis. J Alzheimers Dis 2015; 44: 385–96.
72 Wang Z, Wei X, Yang J, et al. Chronic exposure to aluminum and risk of Alzheimer’s disease: A meta-analysis. Neurosci Lett 2016; 610: 200–6.
73 Zhang Y, Chen J, Qiu J, Li Y, Wang J, Jiao J. Intakes of fish and PUFAs and mild-to-severe cognitive impairment risks: a dose-response meta-analysis of 21 cohort studies. Am J Clin Nutr 2015; published online Dec 30. DOI:10.3945/ajcn.115.124081.
74 Zhou J, Yu J-T, Wang H-F, et al. Association between stroke and Alzheimer’s disease: systematic review and meta-analysis. J Alzheimers Dis 2015; 43: 479–89.
75 Blondell SJ, Hammersley-Mather R, Veerman J. Does physical activity prevent cognitive decline and dementia?: A systematic review and meta-analysis of longitudinal studies. BMC Public Health 2014; 14: 510.
76 Kuiper JS, Zuidersma M, Oude Voshaar RC, et al. Social relationships and risk of dementia: A systematic review and meta-analysis of longitudinal cohort studies. Ageing Res Rev 2015; 22: 39–57.
77 Levi Marpillat N, Macquin-Mavier I, Tropeano A-I, Bachoud-Levi A-C, Maison P. Antihypertensive classes, cognitive decline and incidence of dementia: a network meta-analysis. J Hypertens 2013; 31: 1073–82.
78 Loef M, Walach H. Midlife obesity and dementia: meta-analysis and adjusted
24
forecast of dementia prevalence in the United States and China. Obesity (Silver Spring) 2013; 21: E51–5.
79 Pedditizi E, Peters R, Beckett N. The risk of overweight/obesity in mid-life and late life for the development of dementia: a systematic review and meta-analysis of longitudinal studies. Age Ageing 2016; 45: 14–21.
80 Russ TC, Batty GD, Hearnshaw GF, Fenton C, Starr JM. Geographical variation in dementia: Systematic review with meta-analysis. Int J Epidemiol 2012; 41: 1012–32.
81 Santangeli P, Di Biase L, Bai R, et al. Atrial fibrillation and the risk of incident dementia: A meta-analysis. Hear Rhythm 2012; 9: 1761–8.e2.
82 Shen T, Lv J, Wang L, Wang W, Zhang D. Association between tooth loss and dementia among older people: a meta-analysis. Int J Geriatr Psychiatry 2015; published online Dec 7. DOI:10.1002/gps.4396.
83 Ungprasert P, Wijarnpreecha K, Thongprayoon C. Rheumatoid arthritis and the risk of dementia: A systematic review and meta-analysis. Neurol India 2016; 64: 56–61.
84 Wu S, Ding Y, Wu F, Li R, Hou J, Mao P. Omega-3 fatty acids intake and risks of dementia and Alzheimer’s disease: A meta-analysis. Neurosci Biobehav Rev 2015; 48: 1–9.
85 Zhong G, Wang Y, Zhang Y, Zhao Y. Association between Benzodiazepine Use and Dementia: A Meta-Analysis. PLoS One 2015; 10: e0127836.
86 Zhong G, Wang Y, Zhang Y, Guo JJ, Zhao Y. Smoking Is Associated with an Increased Risk of Dementia: A Meta-Analysis of Prospective Cohort Studies with Investigation of Potential Effect Modifiers. PLoS One 2015; 10: e0118333.
87 Aarsland D, Sardahaee FS, Anderssen S, Ballard C. Is physical activity a potential preventive factor for vascular dementia? A systematic review. Aging Ment Health 2010; 14: 386–95.
88 Beydoun MA, Beydoun HA, Wang Y. Obesity and central obesity as risk factors for incident dementia and its subtypes: a systematic review and meta-analysis. Obes Rev 2008; 9: 204–18.
25
89 Sharp SI, Aarsland D, Day S, Sønnesyn H, Alzheimer’s Society Vascular Dementia Systematic Review Group, Ballard C. Hypertension is a potential risk factor for vascular dementia: systematic review. Int J Geriatr Psychiatry 2011; 26: 661–9.
26
Figure 1. Flow chart of literature search
27
2543 articles reviewed by title screening
150 articles reviewed by abstract screening
120 articles reviewed by full text screening
43 eligible articles published until January 16, 2016
2393 articles were excluded
693 were treatment studies504 had outcomes other than risk for dementia457 were articles about genetic epidemiology318 were editorials or narrative reviews200 were imaging, diagnostic or prognostic studies87 were articles about biomarkers68 were incidence or prevalence studies37 were articles about health economics28 were methodological papers1 was a retracted meta-analysis
30 articles were excluded
28 had outcomes other than risk for dementia1 was primary study1 was article about biomarkers
77 articles were excluded
40 were not the largest meta-analyses investigating a risk factor32 were systematic reviews without a quantitative synthesis3 were pooled analyses without a systematic review of literature1 was a meta-analysis including less than 3 component studies1 was a meta-analysis with poor quality of reporting and statistical analysis
Table 1. Characteristics and quantitative synthesis of the 76 eligible meta-analyses of environmental risk factors for Alzheimer’s disease, all
types of dementia and vascular dementia
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Alzheimer’s disease
Anstey, 200949Alcohol
drinking
Light or moderate
drinkers vs.
Never drinkers
709/11784 6 RR 0.72 (0.61-0.86) 2.0 × 10-4 0.44-1.18 56.4 No/No Weak
Anstey, 201150 Midlife BMIUnderweight vs.
normal weight1041/7218 3 RR 1.78 (0.84-3.75) 0.133 10-4-104 71.2 No/No NS
Beckett, 201551Physical
activity
High level vs.
Low level1358/18968 9 HR 0.62 (0.52-0.72) 5.0 × 10-9 0.51-0.75 0 Yes/No
Highly
suggestive
Cataldo, 201052 SmokingEver vs. Never
smokers6927/325054 43 RR 1.05 (0.91-1.20) 0.526 0.52-2.10 64.8 No/Yes NS
Chang-Quan,
201153
Anti-
hypertensive
drugs
Exposed vs. Not
exposed906/20838 7 RR 0.90 (0.79-1.03) 0.133 0.76-1.08 0 No/No NS
da Silva, 201354Depression at
any age
Diseased vs. Non-
diseased5101/41055 25 RR 1.77 (1.48-2.13) 6.0 × 10-10 0.86-3.66 69.6 Yes/Yes
Highly
suggestive
Diniz, 201336Late-life
depression
Diseased vs. Non-
diseased3358/30111 16 RR 1.65 (1.42-1.92) 4.8 × 10-11 1.36-1.99 2.2 No/No Convincing
28
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Garcia, 200855
Low-frequency
electromagneti
c fields
Exposed vs. Not
exposed3238/10446519 25 RR 1.74 (1.37-2.21) 5.9 × 10-6 0.77-3.91 55.2 Yes/Yes Suggestive
Gudala, 201356
Type 2
diabetes
mellitus
Diseased vs. Not
diseased3537/529160 21 RR 1.54 (1.39-1.72) 3.1 × 10-15 1.37-1.73 0 No/No Convincing
Kim, 201557 Caffeine intakeHigh intake vs.
Low intake590/6280 5 RR 0.78 (0.50-1.22) 0.275 0.17-3.53 71 No/No NS
Li, 201258Dietary intake
of vitamin C
High intake vs.
Low intake1043/13468 6 RR 0.85 (0.74-0.96) 0.011 0.71-1.01 0 No/No Weak
Li, 201258Dietary intake
of vitamin E
High intake vs.
Low intake1205/14509 7 RR 0.80 (0.67-0.95) 0.011 0.52-1.24 46.7 No/No Weak
Li, 201258Dietary intake
of β carotene
High intake vs.
Low intake801/9445 5 RR 0.92 (0.76-1.13) 0.435 0.59-1.45 18.4 No/No NS
Maheshwari,
201459
Chlamydia
pneumoniae
infection
Diseased vs. Not
diseased282/226 11 OR 6.00 (1.93-18.66) 2.0 × 10-3 0.19-193.81 73 No/Yes Weak
Maheshwari,
201459
Spirochetal
infection
Diseased vs. Not
diseased300/255 13 OR 10.65 (3.40-33.42) 5.0 × 10-5 0.41-279.54 51.6 No/No Weak
Meng, 201260 EducationLow level vs.
High level2769/51532 16 RR 1.82 (1.36-2.43) 5.5 × 10-5 0.55-6.05 90.1 No/No Suggestive
29
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Meng, 201461 Midlife BMIObese vs. Normal
weight1003/16709 5 RR 1.81 (1.22-2.69) 3.0 × 10-3 0.52-6.29 63.7 No/No Weak
O’Brien, 201462 HRT Ever vs. Never 1156/13210 9 HR 0.88 (0.66-1.16) 0.361 0.41-1.87 47.7 No/No NS
Perry, 201563Mild traumatic
brain injury
Exposed vs. Not
exposed7158/21603 19 OR 1.40 (1.03-1.90) 0.034 0.39-4.98 85.2 No/No Weak
Peters, 201464
Calcium
channel
blockers
Ever vs. Never 548/8886 4 RR 0.80 (0.54-1.17) 0.242 0.17-3.75 61.1 No/No NS
Power, 201165 HypertensionDiseased vs. Not
diseased1255/25297 13 RR 0.97 (0.81-1.17) 0.745 0.57-1.64 45.7 No/Yes NS
Richardson,
201366Statins
Exposed vs. Not
exposed3785/762841 13 RR 0.72 (0.59-0.89) 1.9 × 10-3 0.39-1.35 54.7 Yes/Yes Weak
Seitz, 201167General
anesthesia
Exposed vs. Not
exposed2122/5716 15 OR 1.05 (0.93-1.19) 0.448 0.91-1.21 0 No/No NS
Steel, 201568Herpesviridae
infection
Diseased vs. Not
diseased1330/1565 33 OR 1.38 (1.14-1.65) 7.3 × 10-4 0.86-2.21 20.3 No/Yes Suggestive
Terracciano,
201469Agreeableness
High level vs.
Low level382/2960 3 HR 0.88 (0.79-0.98) 0.019 0.43-1.78 0 No/No Weak
Terracciano,
201469
Conscientiousn
ess
High level vs.
Low level382/2960 3 HR 0.77 (0.69-0.86) 1.5 × 10-6 0.38-1.54 0 Yes/Yes Weak
30
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Terracciano,
201469Extraversion
High level vs.
Low level382/2960 3 HR 0.95 (0.82-1.11) 0.530 0.23-3.91 39.8 No/No NS
Terracciano,
201469Neuroticism
High level vs.
Low level607/4447 5 HR 1.33 (1.21-1.45) 1.9 × 10-9 1.14-1.54 0 No/Yes Weak
Terracciano,
201469Openness
High level vs.
Low level382/2960 3 HR 0.86 (0.77-0.96) 8.3 × 10-3 0.41-1.79 0 No/No Weak
Virk, 201570 Antacid drugs Ever vs. Never 939/5371 9 OR 0.96 (0.77-1.21) 0.747 0.73-1.26 0 No/No NS
Wang, 201571 Aspirin Ever vs. Never 2263/16393 11 RR 0.77 (0.63-0.95) 0.014 0.42-1.42 55.5 Yes/No Weak
Wang, 201571 Corticosteroids Ever vs. Never 351/4023 3 RR 0.62 (0.26-1.46) 0.277 10-4-103 38.3 No/Yes NS
Wang, 201571Non-aspirin
NSAIDsEver vs. Never 2108/15940 9 RR 0.65 (0.49-0.86) 2.3 × 10-3 0.29-1.45 59.1 Yes/No Weak
Wang, 201571 NSAIDs Ever vs. Never 53372/228119 16 RR 0.74 (0.64-0.86) 6.9 × 10-5 0.45-1.22 70 No/No Suggestive
Wang, 201672 AluminumExposed vs. Not
exposed1383/9184 8 OR 1.72 (1.33-2.21) 3.1 × 10-5 1.16-2.54 6.2 No/No Suggestive
Zhang, 201543 CancerDiseased vs. Not
diseased4635/40251 7 HR 0.62 (0.53-0.74) 4.6 × 10-8 0.50-0.78 0 Yes/No
Highly
suggestive
Zhang, 201573Dietary DHA
intake
High intake vs.
Low intake535/6104 3 RR 0.55 (0.24-1.23) 0.145 10-5-104 90.5 No/Yes NS
Zhang, 201573 Fish intakeHigh intake vs.
Low intake915/21026 5 RR 0.88 (0.79-0.98) 0.022 0.63-1.22 63.4 No/No Weak
31
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Zhou, 201574 StrokeDiseased vs. Not
diseased952/13778 6 HR 1.59 (1.25-2.02) 1.7 × 10-4 1.13-2.23 0 No/No Weak
All types of dementia
Anstey, 200949Alcohol
drinking
Light or moderate
drinkers vs.
Never drinkers
1016/16389 7 RR 0.74 (0.61-0.91) 3.9 × 10-3 0.43-1.28 52.6 No/Yes Weak
Blondell, 201475Physical
activity
High level vs.
Low level3845/35974 21 RR 0.76 (0.66-0.86) 1.9 × 10-5 0.49-1.15 68.9 Yes/Yes Suggestive
da Silva, 201354Depression at
any age
Diseased vs. Not
diseased25106/416385 33 RR 1.99 (1.84-2.16) 8.0 × 10-62 1.65-2.40 27.8 No/No Convincing
da Silva, 201354Early-life
depression
Diseased vs. Not
diseased3538/24845 9 RR 1.63 (1.27-2.11) 1.5 × 10-4 1.01-2.64 16.2 No/No Suggestive
Diniz, 201336Late-life
depression
Diseased vs. Not
diseased4957/46396 25 RR 1.85 (1.67-2.05) 3.1 × 10-32 1.66-2.06 0 No/No Convincing
Gudala, 201356
Type 2
diabetes
mellitus
Diseased vs. Not
diseased15707/1125450 22 RR 1.60 (1.43-1.79) 5.4 × 10-17 1.05-2.44 72.3 No/No
Highly
suggestive
Kim, 201557 Caffeine intakeHigh intake vs.
Low intake905/7535 5 RR 0.72 (0.34-1.51) 0.385 0.05-9.50 75.7 No/No NS
Kuiper, 201576Frequency of
social contacts
Low level vs.
High level1122/14640 8 RR 1.57 (1.32-1.85) 1.9 × 10-7 1.27-1.93 0 No/No Convincing
32
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Kuiper, 201576 LonelinessHigh level vs.
Low level280/2972 3 RR 1.58 (1.19-2.09) 1.5 × 10-3 0.25-9.78 0 No/No Weak
Kuiper, 201576
Satisfaction
with social
network
Low level vs.
High level985/5222 4 RR 1.25 (0.96-1.62) 0.103 0.46-3.36 50.3 No/No NS
Kuiper, 201576Social network
size
Low level vs.
High level1059/6691 5 RR 1.17 (0.92-1.48) 0.197 0.56-2.44 64 No/No NS
Kuiper, 201576Social
participation
Low level vs.
High level589/7125 6 RR 1.41 (1.13-1.75) 2.0 × 10-3 0.85-2.34 31.2 Yes/Yes Weak
Levi Marpillat,
201377
Anti-
hypertensive
drugs
Ever vs. Never 86422/1436995 11 HR 0.84 (0.75-0.94) 1.7 × 10-3 0.60-1.16 73.4 Yes/No Weak
Loef, 201378 Midlife BMIObese vs. Normal
weight1914/28405 5 RR 1.91 (1.40-2.62) 5.1 × 10-5 0.74-4.93 53.5 No/No Suggestive
Meng, 201260 EducationLow level vs.
High level8739/78504 23 RR 1.88 (1.51-2.33) 1.2 × 10-8 0.69-5.14 89.6 No/Yes Suggestive
Pedditizi, 201679 Late-life BMIObese vs. Normal
weight1053/7400 4 RR 0.83 (0.74-0.94) 3.0 × 10-3 0.64-1.09 0 No/No Weak
Pedditizi, 201679 Late-life BMIOverweight vs.
Normal weight1885/12394 5 RR 0.88 (0.76-1.02) 0.094 0.56-1.38 54.3 No/No NS
33
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Pedditizi, 201679 Midlife BMIOverweight vs.
Normal weight1252/12792 4 RR 1.10 (0.99-1.22) 0.076 0.87-1.38 0 No/No NS
Perry, 201563Mild traumatic
brain injury
Exposed vs. Not
exposed7798/22564 23 OR 1.35 (1.01-1.78) 0.040 0.38-4.75 85 No/No Weak
Richardson,
201366Statins
Ever vs. never
users37798/4325018 12 RR 0.83 (0.76-0.91) 1.3 × 10-4 0.66-1.04 63.2 Yes/No Suggestive
Russ, 201280 Rural livingExposed vs. Not
exposed832/10504 4 RR 1.12 (0.75-1.69) 0.583 0.18-7.03 83.7 No/Yes NS
Santangeli, 201281Atrial
fibrillation
Diseased vs. Not
diseased5301/68586 9 HR 1.36 (1.12-1.65) 2.0 × 10-3 0.82-2.26 52.5 No/No Weak
Shen, 201582 Tooth loss High vs. Low 2378/17439 10 RR 1.56 (1.25-1.96) 1.1 × 10-4 0.89-2.75 44.8 No/Yes Weak
Ungprasert,
201683
Rheumatoid
arthritis
Diseased vs. Not
diseased19088/3044201 4 RR 1.58 (1.04-2.40) 0.031 0.25-10.14 93.4 No/No Weak
Wu, 201584 Fish intakeHigh intake vs.
Low intake1013/20574 5 RR 0.79 (0.62-1.01) 0.060 0.42-1.48 29.1 Yes/No NS
Zhong, 201585Benzodiazepin
es use
Ever vs. Never
users11741/29981 5 RR 1.49 (1.30-1.72) 2.7 × 10-8 1.03-2.17 35.1 No/No Convincing
Zhong, 201586 SmokingEver vs. Never
smokers14944/930070 27 RR 1.13 (1.05-1.22) 1.2 × 10-3 0.87-1.47 47.1 No/Yes Weak
Vascular dementia
34
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Aarsland, 201087Physical
activity
High level vs.
Low level374/10108 5 RR 0.62 (0.42-0.92) 0.017 0.19-2.07 55.7 No/No Weak
Anstey, 200949Alcohol
drinking
Light or moderate
drinkers vs.
Never drinkers
151/8969 4 RR 0.75 (0.57-0.98) 0.037 0.38-1.45 5.2 No/No Weak
Beydoun, 200888 Midlife BMIObese vs. Normal
weight222/14017 3 RR 1.71 (0.47-6.26) 0.419 10-7-107 89.1 No/No NS
Chang-Quan,
201153
Anti-
hypertensive
drugs
Ever vs. Never 254/18558 4 RR 0.64 (0.42-0.98) 0.040 0.14-2.96 46.8 No/No Weak
da Silva, 201354Depression at
any age
Diseased vs. Not
diseased227/12929 4 RR 2.92 (1.87-4.56) 2.5 × 10-6 1.10-7.78 0 No/No Weak
Diniz, 201336Late-life
depression
Diseased vs. Not
diseased316/16250 5 OR 2.52 (1.77-3.59) 2.8 × 10-7 1.40-4.54 1.1 No/No Weak
Gudala, 201356
Type 2
diabetes
mellitus
Diseased vs. Not
diseased1396/875524 14 RR 2.28 (1.94-2.66) 1.1 × 10-24 1.91-2.71 0 No/No Convincing
Meng, 201260 EducationLow level vs.
High level379/7334 3 RR 2.75 (2.19-3.45) 2.1 × 10-18 0.63-11.96 0 No/No Weak
Sharp, 201189 HypertensionDiseased vs. Not
diseased425/7698 6 HR 1.59 (1.20-2.11) 1.4 × 10-3 0.78-3.21 36.3 No/No Weak
35
Reference Risk factorLevel of
comparison
Total number
of
cases/controls
Number of
primary
studies
Effect
size
metric
Random effects
summary effect
size (95% CI)
P random 95% PI I2
Small-study
effects/Excess
statistical
significance
Level of
evidence
Zhong, 201586 SmokingEver vs. Never
smokers1406/885388 8 RR 1.26 (1.05-1.50) 0.013 0.79-2.00 43.9 No/No Weak
BMI: body mass index, CI: confidence intervanl, DHA: docosahexaenoic acid, HR: hazard ratio, HRT: hormone replacement therapy, NSAIDs: non-steroid anti-inflammatory drugs, OR: odds ratio, PI: prediction interval, RR: risk ratio
36
Table 2. Assessment of the statistically significant environmental risk factors for dementia (Alzheimer’s disease, all types of dementia, vascular dementia)
Level of evidence Criteria Alzheimer’s disease All types of dementia Vascular dementia
Convincing
>1000 cases, p<10-6, I2<50%, 95% PI excluding the null value, no small-study effects and excess significance bias
Late-life depression, type 2 diabetes mellitus
Benzodiazepines use, depression at any age, frequency of social contacts, late-life depression
Type 2 diabetes mellitus
Highly suggestive
>1000 cases, p<10-6, largest study with a statistically significant effect
Cancer, depression at any age, physical activity Type 2 diabetes mellitus None
Suggestive>1000 cases, p<10-3
Aluminum, education, herpesviridae infection, Low-frequency electromagnetic fields, NSAIDs
Early-life depression, education, physical activity, midlife BMI (Obese vs. Normal weight), statins
None
WeakThe rest associations with p<0.05
Alcohol drinking, dietary intake of vitamin C, dietary intake of vitamin E, chlamydia pneumoniae infection, spirochetal infection, midlife BMI (Obese vs. Normal weight), mild traumatic brain injury, statins, agreeableness, conscientiousness, neuroticism, openness, aspirin, non-aspirin NSAIDs, fish intake, stroke
Alcohol drinking, anti-hypertensive drugs, atrial fibrillation, loneliness, social participation, late-life BMI (Obese vs. Normal weight), mild traumatic brain injury, tooth loss, rheumatoid arthritis, smoking
Alcohol drinking, anti-hypertensive drugs, depression at any age, hypertension, late-life depression, education, physical activity, smoking
37
Table 3. Sensitivity analysis limited to prospective cohort studies for associations with convincing and highly suggestive evidence in the main analysis
References Risk factor N of studies
Number of cases/controls
Effect size
metric
Random-effects summary effect size (95% CI)
P random 95% PI I2
Small-study effects/Excess
statistical significance
Level of evidence
Alzheimer’s diseaseDiniz, 201336 Late-life depression 15 3348/29951 RR 1.64 (1.40-1.92) 1.4 × 10-9 1.27-2.12 7.5 No/No Convincing
Gudala, 201356 Type 2 diabetes mellitus 21 3537/529160 RR 1.54 (1.39-1.72) 3.1 × 10-15 1.37-1.73 0 No/No Convincing
da Silva, 201354 Depression at any age 15 1461/24937 RR 1.72 (1.39-2.13) 4.9 × 10-7 0.90-3.31 57.5 Yes/Yes Highly suggestive
Zhang, 201543 Cancer 6 1354/27127 HR 0.63 (0.53-0.75) 1.0 × 10-7 0.49-0.80 0 Yes/No Highly suggestive
Beckett. 201551 Physical activity 9 1358/18968 HR 0.62 (0.52-0.72) 5.0 × 10-9 0.51-0.75 0 Yes/No Highly suggestive
All types of dementiaZhong, 201585 Benzodiazepines use 5 11741/29981 RR 1.49 (1.30-1.72) 2.7 × 10-8 1.03-2.17 35.1 No/No Convincing
da Silva, 201354 Depression at any age 23 2781/29578 RR 1.86 (1.61-2.14) 2.6 × 10-17 1.27-2.71 26 No/No Convincing
Kuiper, 201576 Frequency of social contacts 8 1122/14640 RR 1.57 (1.32-1.85) 1.9 × 10-7 1.27-1.93 0 No/No Convincing
Diniz, 201336 Late-life depression 22 4782/45306 RR 1.83 (1.65-2.03) 3.3 × 10-29 1.63-2.05 0 No/No Convincing
Gudala, 201356 Type 2 diabetes mellitus 22 15707/1125450 RR 1.60 (1.43-1.79) 5.4 × 10-17 1.05-2.44 72.3 No/No Highly
suggestiveVascular dementia
Gudala, 201356 Type 2 diabetes mellitus 14 1396/875524 RR 2.28 (1.94-2.66) 1.1 × 10-24 1.91-2.71 0 No/No Convincing
38
HR: hazard ratio, PI: prediction interval, RR: risk ratio
39
Supplementary Table 2. Quality assessments of primary studies based on Newcastle-Ottawa scale
Reference Risk factor OutcomeHigh quality
(NOS score = 9)
Moderate quality
(NOS score = 7 or 8)
Low quality
(NOS score < 7)
Diniz, 201336
Late-life depression AD 9 6 1
Late-life depression All types of dementia 11 12 2
Late-life depression VaD 3 2 0
Kim, 201557Caffeine intake AD 0 4 1
Caffeine intake All types of dementia 1 0 4
Richardson, 201366Statins AD 3 10 0
Statins All types of dementia 2 9 1
Seitz, 201167 General anesthetics AD 4 1 10
Wang, 201571
Aspirin AD 0 6 5
Corticosteroids AD 0 1 2
Non-aspirin NSAIDs AD 0 6 3
NSAIDs AD 1 10 5
da Silva, 201354
Depression at any age All types of dementia 6 17 10
Early-life depression All types of dementia 2 7 0
Depression at any age VaD 1 3 0
Gudala, 201356Type 2 diabetes mellitus All types of dementia 16 4 0
Type 2 diabetes mellitus VaD 11 2 0
Wu, 201584 Fish intake All types of dementia 0 3 2
40
Reference Risk factor OutcomeHigh quality
(NOS score = 9)
Moderate quality
(NOS score = 7 or 8)
Low quality
(NOS score < 7)
Zhong, 201585 Benzodiazepines use All types of dementia 0 4 1
Zhong, 201586Smoking All types of dementia 1 12 14
Smoking VaD 1 5 2
AD: Alzheimer’s disease, NOS: Newcastle-Ottawa scale, NSAIDs: non-steroid anti-inflammatory drugs, VaD: vascular dementia
41
Supplementary Table 3. Heterogeneity estimates, bias assessment and largest study effect size across the 76 eligible meta-analyses of
environmental risk factors for Alzheimer’s disease, all types of dementia and vascular dementia
Reference Risk factor
Effect
size
metric
Largest study effect
size (95% CI)SE I2 (%)
Egger test p-
value
Observed
significant studies
Expected
significant studies
Excess significance
test p-value
Alzheimer’s disease
Anstey, 200949 Alcohol drinking RR 0.63 (0.55-0.72) 0.069 56.4 0.363 2 3.98 NP
Anstey, 201150Midlife BMI
(Underweight vs. Normal weight)RR 3.43 (1.90-6.20) 0.302 71.2 0.121 1 2.97 NP
Beckett, 201551 Physical activity HR 0.69 (0.50-0.96) 0.168 0 0.016 6 5.43 0.697
Cataldo, 201052 Smoking RR 0.88 (0.73-1.10) 0.105 64.8 0.702 13 4.56 NP
Chang-Quan, 201153 Anti-hypertensive drugs RR 0.88 (0.68-1.13) 0.130 0 0.161 0 0.86 NP
da Silva, 201354 Depression at any age RR 1.19 (1.07-1.32) 0.054 69.6 0.009 16 4.6 1.0 × 10-8
Diniz, 201336 Late-life depression RR 1.43 (1.05-1.96) 0.159 2.2 0.804 7 6.91 0.964
Garcia, 200855 Low-frequency electromagnetic fields RR 1.20 (1.00-1.40) 0.086 55.2 0.001 11 3.25 3.87 × 10-6
Gudala, 201356 Type 2 diabetes mellitus RR 1.60 (1.29-1.98) 0.109 0 0.629 9 15.04 NP
Kim, 201557 Caffeine intake RR 0.69 (0.50-0.96) 0.166 71 0.406 2 2.13 NP
Li, 201258 Dietary intake of vitamin C RR 0.83 (0.68-1.01) 0.101 0 0.875 1 1.45 NP
Li, 201258 Dietary intake of vitamin E RR 0.74 (0.62-0.88) 0.089 46.7 0.869 3 3.42 NP
Li, 201258 Dietary intake of β-carotene RR 0.81 (0.63-1.03) 0.125 18.4 0.901 0 1.34 NP
42
Reference Risk factor
Effect
size
metric
Largest study effect
size (95% CI)SE I2 (%)
Egger test p-
value
Observed
significant studies
Expected
significant studies
Excess significance
test p-value
Maheshwari, 201459 Chlamydia pneumoniae infection OR 1.10 (0.51-2.38) 0.393 73 0.264 4 0.59 4.62 × 10-6
Maheshwari, 201459 Spirochetal infection OR 4.47 (1.92-10.40) 0.431 51.6 0.235 6 6.4 NP
Meng, 201260 Education RR 3.83 (3.16-4.63) 0.097 90.1 0.442 11 16 NP
Meng, 201461Midlife BMI
(Obese vs. Normal weight)RR 1.68 (1.21-2.33) 0.167 63.7 0.473 2 3.97 NP
O’Brien, 201462 HRT HR 0.80 (0.58-1.09) 0.161 47.7 0.978 1 1.95 NP
Perry, 201563 Mild traumatic brain injury OR 4.03 (3.27-4.96) 0.106 85.2 0.028 8 18.88 NP
Peters, 201464 Calcium channel blockers RR 1.08 (0.81-1.45) 0.149 61.1 0.163 1 0.30 0.189
Power, 201165 Hypertension RR 1.05 (0.76-1.47) 0.168 45.7 0.784 3 0.74 0.007
Richardson, 201366 Statins RR 1.14 (0.85-1.53) 0.150 54.7 0.085 8 2.79 4.40 × 10-4
Seitz, 201167 General anesthesia OR 1.18 (0.98-1.42) 0.095 0 0.159 0 1.98 NP
Steel, 201568 Herpesviridae infection OR 1.22 (1.01-1.48) 0.097 20.3 0.830 6 2.51 0.022
Terracciano, 201469 Agreeableness HR 0.90 (0.77-1.06) 0.083 0 0.358 0 0.28 NP
Terracciano, 201469 Conscientiousness HR 0.81 (0.69-0.95) 0.080 0 0.072 3 0.68 1.39 × 10-3
Terracciano, 201469 Extraversion HR 1.09 (0.91-1.31) 0.093 39.8 0.478 0 0.24 NP
Terracciano, 201469 Neuroticism HR 1.19 (1.01-1.41) 0.085 0 0.156 5 0.82 4.60 × 10-7
Terracciano, 201469 Openness HR 0.91 (0.76-1.08) 0.090 0 0.896 1 0.25 0.120
Virk, 201570 Antacid drugs OR 0.91 (0.65-1.26) 0.169 0 0.791 0 0.64 NP
Wang, 201571 Aspirin RR 0.84 (0.63-1.11) 0.144 55.5 0.068 2 2.43 NP
Wang, 201571 Corticosteroids RR 1.03 (0.49-2.15) 0.377 38.3 0.373 1 0.16 0.029
Wang, 201571 Non-aspirin NSAIDs RR 1.19 (0.87-1.62) 0.159 59.1 0.017 4 2.20 0.160
43
Reference Risk factor
Effect
size
metric
Largest study effect
size (95% CI)SE I2 (%)
Egger test p-
value
Observed
significant studies
Expected
significant studies
Excess significance
test p-value
Wang, 201571 NSAIDs RR 0.76 (0.68-0.85) 0.057 70 0.117 10 8.80 0.545
Wang, 201672 Aluminum OR 1.70 (1.20-2.60) 0.197 6.2 0.931 3 6.18 NP
Zhang, 201543 Cancer HR 0.64 (0.50-0.81) 0.123 0 0.037 3 5.79 NP
Zhang, 201573 Fish iintake RR 0.93 (0.91-0.95) 0.011 63.4 0.174 3 0.41 2.0 × 10-5
Zhang, 201573 Dietary DHA intake RR 1.10 (0.93-1.31) 0.087 90.5 0.110 2 0.31 0.001
Zhou, 201574 Stroke HR 1.20 (0.77-1.89) 0.229 0 0.924 2 1.31 0.498
All types of dementia
Anstey, 200949 Alcohol drinking RR 0.80 (0.65-0.99) 0.107 52.6 0.634 5 1.86 0.007
Blondell, 201475 Physical activity RR 1.04 (0.98-1.10) 0.029 68.9 2.92 × 10-5 8 1.23 1.00 × 10-8
da Silva, 201354 Depression at any age RR 2.18 (2.08-2.28) 0.023 27.8 0.387 19 27.99 NP
da Silva, 201354 Early-life depression RR 1.58 (1.06-2.37) 0.205 16.2 0.484 4 6.26 NP
Diniz, 201336 Late-life depression RR 1.72 (1.38-2.13) 0.111 0 0.736 14 16.37 NP
Gudala, 201356 Type 2 diabetes mellitus RR 1.62 (1.49-1.77) 0.044 72.3 0.081 14 20.23 NP
Kim, 201557 Caffeine intake RR 1.12 (0.66-1.91) 0.271 75.7 0.600 1 0.61 0.590
Kuiper, 201576 Frequency of social contacts RR 1.67 (1.25-2.23) 0.148 0 0.961 4 6.33 NP
Kuiper, 201576 Loneliness RR 1.54 (1.08-2.19) 0.180 0 0.120 1 1.75 NP
Kuiper, 201576 Satisfaction with social network RR 1.30 (1.06-1.59) 0.103 50.3 0.932 1 2.03 NP
Kuiper, 201576 Social network size RR 0.99 (0.95-1.03) 0.021 64 0.144 1 0.25 0.128
Kuiper, 201576 Social participation RR 1.18 (1.07-1.31) 0.052 31.2 0.080 3 0.83 0.010
Levi Marpillat,
201377Anti-hypertensive drugs HR 0.94 (0.91-0.97) 0.016 73.4 0.099 4 2.61 0.326
44
Reference Risk factor
Effect
size
metric
Largest study effect
size (95% CI)SE I2 (%)
Egger test p-
value
Observed
significant studies
Expected
significant studies
Excess significance
test p-value
Loef, 201378Midlife BMI
(Obese vs. Normal weight)RR 1.74 (1.34-2.26) 0.133 53.5 0.415 3 4.88 NP
Meng, 201260 Education RR 1.29 (1.17-1.43) 0.051 89.6 0.296 16 8.63 0.001
Pedditizi, 201679Late-life BMI
(Obese vs. Normal weight)RR 0.81 (0.69-0.94) 0.079 0 0.127 1 1.48 NP
Pedditizi, 201679Late-life BMI
(Overwegith vs. Normal weight)RR 0.81 (0.72-0.91) 0.060 54.3 0.873 1 2.18 NP
Pedditizi, 201679Midlife BMI
(Overweight vs. Normal weight)RR 1.17 (0.99-1.38) 0.085 0 0.794 0 1.19 NP
Perry, 201563 Mild traumatic brain injury OR 4.03 (3.27-4.96) 0.106 85 0.012 9 22.86 NP
Richardson, 201366 Statins RR 0.88 (0.85-0.91) 0.017 63.2 0.051 8 5.65 0.174
Russ, 201280 Rural living RR 0.94 (0.79-1.13) 0.091 83.7 0.412 2 0.29 1.06 × 10-3
Santangeli, 201281 Atrial fibrillation HR 1.36 (1.13-1.63) 0.093 52.5 0.832 5 5.69 NP
Shen, 2015 Tooth loss RR 1.10 (0.89-1.36) 0.108 44.8 0.110 5 1.15 1.3 × 10-4
Ungprasert, 2016 Rheumatoid arthritis RR 2.02 (1.83-2.23) 0.050 93.4 0.810 2 2.79 NP
Wu, 201584 Fish intake RR 0.95 (0.76-1.19) 0.114 29.1 0.051 1 0.33 0.234
Zhong, 201585 Benzodiazepines use RR 1.60 (1.37-1.87) 0.079 35.1 0.516 5 4.65 0.541
Zhong, 201586 Smoking RR 1.07 (0.96-1.19) 0.055 47.1 0.641 8 3.39 7.40 × 10-3
Vascular dementia
Aarsland, 201087 Physical activity RR 0.85 (0.65-1.13) 0.141 55.7 0.115 1 0.59 0.570
Anstey, 200949 Alcohol drinking RR 0.79 (0.50-1.25) 0.234 5.2 0.616 1 0.46 0.399
45
Reference Risk factor
Effect
size
metric
Largest study effect
size (95% CI)SE I2 (%)
Egger test p-
value
Observed
significant studies
Expected
significant studies
Excess significance
test p-value
Bedoun, 200888 Obesity RR 4.95 (2.98-8.43) 0.265 89.1 0.475 1 3.00 NP
Chang-Quan, 201153 Anti-hypertensive drugs RR 0.66 (0.45-0.97) 0.196 46.8 0.363 2 1.69 0.720
da Silva, 201354 Depression at any age RR 2.41 (1.22-4.52) 0.334 0 0.665 2 3.01 NP
Diniz, 201336 Late-life depression OR 1.79 (0.99-3.22) 0.301 1.1 0.586 2 3.12 NP
Gudala, 201356 Type 2 diabetes mellitus RR 2.00 (1.50-2.66) 0.146 0 0.592 10 11.07 NP
Meng, 201260 Education RR 2.95 (2.14-4.07) 0.164 0 0.570 3 2.99 0.968
Sharp, 201189 Hypertension OR 1.53 (1.10-2.13) 0.169 36.3 0.730 3 2.62 0.754
Zhong, 201586 Smoking RR 1.02 (0.84-1.25) 0.101 43.9 0.326 2 0.42 0.012
HR: hazard ratio, OR: odds ratio, RR: relative risk
46