Available via license: CC BY-NC 4.0
Content may be subject to copyright.
1
Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
REVIEW
Evidence- based prevention of Alzheimer's disease:
systematic review and meta- analysis of 243
observational prospective studies and 153
randomised controlledtrials
Jin- Tai Yu ,1 Wei Xu ,2 Chen- Chen Tan,2 Sandrine Andrieu,3 John Suckling,4
Evangelos Evangelou,5 An Pan,6 Can Zhang,7 Jianping Jia,8 Lei Feng,9 Ee- Heok Kua,9
Yan- Jiang Wang,10 Hui- Fu Wang,2 Meng- Shan Tan,2 Jie- Qiong Li,2 Xiao- He Hou,2
Yu Wan,2 Lin Tan,2 Vincent Mok,11 Lan Tan,2 Qiang Dong,1 Jacques Touchon,12
Gauthier Serge,13 Paul S Aisen,14 Bruno Vellas15
Cognitive neurology
To cite: Yu J- T, Xu W, Tan
C- C, etal. J Neurol Neurosurg
Psychiatry Epub ahead of
print: [please include Day
Month Year]. doi:10.1136/
jnnp-2019-321913
►Additional material is
published online only. To view
please visit the journal online
(http:// dx. doi. org/ 10. 1136/
jnnp- 2019- 321913).
For numbered affiliations see
end of article.
Correspondence to
Professor Jin- Tai Yu, Department
of Neurology and Institute of
Neurology, Huashan Hospital,
Shanghai Medical College,
Fudan University, Shanghai
200040, China; jintai_ yu@
fudan. edu. cn
J- TY, WX and C- CT contributed
equally.
Received 22 August 2019
Revised 26 February 2020
Accepted 22 April 2020
© Author(s) (or their
employer(s)) 2020. Re- use
permitted under CC BY- NC. No
commercial re- use. See rights
and permissions. Published
by BMJ.
ABSTRACT
Background Evidence on preventing Alzheimer’s
disease (AD) is challenging to interpret due to varying
study designs with heterogeneous endpoints and
credibility. We completed a systematic review and meta-
analysis of current evidence with prospective designs to
propose evidence- based suggestions on AD prevention.
Methods Electronic databases and relevant websites
were searched from inception to 1 March 2019. Both
observational prospective studies (OPSs) and randomised
controlled trials (RCTs) were included. The multivariable-
adjusted effect estimates were pooled by random- effects
models, with credibility assessment according to its risk
of bias, inconsistency and imprecision. Levels of evidence
and classes of suggestions were summarised.
Results A total of 44 676 reports were identified, and
243 OPSs and 153 RCTs were eligible for analysis after
exclusion based on pre- decided criteria, from which
104 modifiable factors and 11 interventions were
included in the meta- analyses. Twenty- one suggestions
are proposed based on the consolidated evidence, with
Class I suggestions targeting 19 factors: 10 with Level
A strong evidence (education, cognitive activity, high
body mass index in latelife, hyperhomocysteinaemia,
depression, stress, diabetes, head trauma, hypertension
in midlife and orthostatic hypotension) and 9 with Level
B weaker evidence (obesity in midlife, weight loss in late
life, physical exercise, smoking, sleep, cerebrovascular
disease, frailty, atrial fibrillation and vitamin C). In
contrast, two interventions are not recommended:
oestrogen replacement therapy (Level A2) and
acetylcholinesterase inhibitors (Level B).
Interpretation Evidence- based suggestions are
proposed, offering clinicians and stakeholders current
guidance for the prevention of AD.
INTRODUCTION
An unequivocal downtrend in the prevalence and
incidence of dementia was recently reported and
associated with earlier population- level investment
(eg, improved education and vascular health),1–3
strengthening the necessity for primary prevention.4
The past few decades have witnessed great global
efforts in updating and upgrading the evidence
on how to prevent Alzheimer’s disease (AD),5 6
accounting for approximately two- thirds of all cases
of dementia and affecting up to 20% of individuals
older than 80 years.7 8 Nevertheless, key issues in
the field are the inconsistency among conclusions
and variable levels of credibility arising from the
wide variety of study designs.9 Two types of studies
are generally regarded as having the greatest impact
on the extant literature: (1) observational prospec-
tive studies (OPSs), which describe temporal rela-
tionships with potential causal links and often use
large samples recruited from community dwellers;
and (2) randomised controlled trials (RCTs), which
possess strong internal validity to infer causality by
testing the effects of specific interventions on the
incidence of AD. Although both approaches are
useful, the major concerns in OPSs are usually the
elusive sources of bias when interpreting the iden-
tified wide- ranging factors, and current RCTs are
often compromised by short follow- up durations,
subjective endpoints, small sample sizes and specific
recruitment criteria with uncertain generalisability.5
Considerable evidence has been generated
regarding AD through OPSs and RCTs. Because it is
almost impossible to conduct RCTs that evaluate all
risk factors of AD, a quantitative depiction of AD’s
prevention 'profile' based on these two complemen-
tary study types is urgently needed for prevention
guidelines that weigh the benefits against the risks.
Deconstructing the bias sources from OPSs will
facilitate the interpretation of credibility ratings
and also guide future research directions. In this
study we consolidated the extant evidence from
both OPSs and RCTs to formulate the levels of
evidence and classes of clinical suggestions for AD
prevention.
METHODS
Search strategy and selection criteria
We followed the recommendations of the Preferred
Reporting Items for Systematic reviews and Meta-
Analyses (PRISMA) 2009 guidelines.10 11 PubMed,
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from
2Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
Cognitive neurology
Figure 1 Flow chart showing literature selection for OPSs (figure1A) and RCTs (figure1B) and map depicting studies eligible for systematic review
(figure1C). A total of 243 OPSs (figure1A) and 153 completed RCTs (figure1B) were finally included. 243 OPSs from 17 countries on four continents
(Europe accounting for 43%, North America 41%, Asia 14% and Latin America 2%) reported the association of 134 modifiable risk factors with risk of
clinical Alzheimer’s- type dementia (83% used all AD, 13% probable AD and 11% pure AD) diagnosed by NINCDS- ADRDA criteria in populations with
various racial backgrounds (68% white, 14% Asian descent, 13% mixed race), sources (84% community, 6% institution, 10% mixed source) and baseline
cognitive statuses (82% free of dementia, 16% cognitively normal, 2% unclear). A total of 153 published RCTs from five continents (North America
accounting for 45%, Europe 36%, Australia 9%, Asia 7% and Latin America 3%) reported the effects of 15 types of interventions on AD (7%), dementia
(16%) and cognitive function (85%) in selected participants, including elderly subjects (37%), high- risk group (35%) or cognitively impaired (28%)
(figure1C). In the pie charts, 1 and 2 show the outcome (all AD=probable or possible AD, or AD with or without VD/CVD, Pure AD=AD without VD or CVD;
A=Alzheimer’s disease, B=Biomarker of AD, C=Cognition, D=Dementia); 3 and 4 show the population source; 5 and 6 show the percentage of studies
from different continents. AD, Alzheimer’s disease; CVD, cerebrovascular disease; OPS, observational prospective study; RCT, randomised controlled trial; VD,
vascular dementia.
EMBASE and CENTRAL were searched using the terms “Alzhei-
mer’s”, “Alzheimer”, “dementia”, and “risk” for OPS and
“Alzheimer”, “cognitive”, “cognition”, “prevent”, and “preven-
tion” for RCT up to 1 March 2019. Bibliographies of relevant
literature and records in Clinicaltrials. gov and AlzRisk data-
base12 were hand- searched in case of omission. The inclusion
criteria were as follows: (1) an OPS exploring the association
between potentially modifiable exposures at baseline and inci-
dent AD independently diagnosed according to the National
Institute of Neurological and Communicative Disorders and
Stroke and the Alzheimer’s Disease and Related Disorders Asso-
ciation (NINCDS- ADRDA) criteria,13 or (2) a RCT targeting the
impact of addressing modifiable risk factors on the incidence
of AD or AD- related clinical endpoints (dementia or cognitive
impairment), and (3) a publication written in English to permit
easy access to the source information of all included articles. The
detailed exclusion criteria are shown in figure 1. Bibliographies
of relevant original studies and systematic reviews were hand-
searched. Literature selection was performed by three pairs of
experienced investigators (JTY, WX, CCT, HFW, MST and JQL)
and any disagreements on inclusion were resolved by consensus
and arbitration by a panel of investigators within the review
team (JTY, WX, CCT, HFW, MST, JQL and Lan Tan).
Data extraction
Pre- designed templates were used to extract the data with refer-
ence to the STROBE statement (https://www. equator- network.
org/ reporting- guidelines/ strobe/). An evidence- based profile of
AD modifiable risk factors was established for better tracing
of bias sources. The multivariable- adjusted risk estimates were
extracted. If these estimates were unavailable, we attempted to
obtain them by contacting the corresponding authors. The strin-
gently performed process comprised three independent steps: (a)
data extraction by three pairs of experienced investigators (JTY,
WX, CCT, HFW, MST and JQL); (b) independent data proof
reading by 10 researchers (JTY, WX, CCT, HFW, MST, JQL,
XHH, YW, Lin Tan and Lan Tan); and (c) addressing discrepan-
cies by consensus and arbitration.
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from
3
Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
Cognitive neurology
Figure 2 Rating levels of evidence and strength of suggestions. Based on the Cochrane Handbook for Systematic Reviews of Interventions for RCTs
and the Newcastle–Ottawa Quality Assessment Scale (NOS) for OPSs, we evaluated the quality of eligible studies. The credibility of each result was then
categorised into four levels: Good (G level), Acceptable (A± level), Susceptible (S± level) and Poor (P level) according to the score combination of three
domains: risk of bias, inconsistency and imprecision. Levels of evidence were summarised, representing the quality of scientific evidence on the basis of
directness of outcome (for RCTs), consistency and quality of data from clinical trials and/or observational studies. Classes of suggestions were made after
weighing the benefits against the risks due to specific interventions. *Factors rated with ‘level C’ evidence were not considered for recommendation in the
present study.
Assessment of study quality and credibility of meta-analyses
The risk of bias tool proposed by Cochrane14 for RCTs and
involving the Newcastle–Ottawa Quality Assessment Scale
(NOS)15 for OPSs were used to evaluate the quality of eligible
studies. The total score for the Cochrane tool or NOS was
regarded as a proxy to assess the overall risk of bias for each
single study. The score for each item evaluated the associated
risk of bias (online supplementary appendix 1). The credibility
of each meta- analysis result was then categorised into four levels:
Good (G level), Acceptable (A± level), Susceptible (S± level)
and Poor (P level) according to the score combination of three
domains: risk of bias,16 inconsistency17 and imprecision18 (online
supplementary appendix 2). In particular, G and A+ levels were
regarded as moderate- to- high credibility.
Levels of evidence and strength of suggestions
Levels of evidence were summarised to represent the quality
of scientific evidence on the basis of directness of outcome for
AD, credibility of meta- analyses and consistency of evidence
from clinical trials and/or observational studies: Level A>Level
B>Level C (based on the evidence level). Classes of recommen-
dations were made after weighing the benefits against the risks
due to specific interventions: Class I (strong recommendation),
Class II (weak recommendation) and Class III (not recom-
mended) (figure 2).
Statistical analyses
The multivariable- adjusted risk estimates and 95% confi-
dence intervals (CI) were log- transformed and combined using
random models (DerSimonian–Laird method).19 Sensitivity anal-
yses excluding odd ratios (ORs) reported by some OPSs were
performed because ORs tend to overestimate the effect size
compared with the relative risk (RR), particularly when the inci-
dence is not small. A 95% prediction interval (PI) was calculated
to better evaluate the precision of the result.20 Heterogeneity
was assessed by Q test and quantified by the I2 metric.21 The
source of heterogeneity was explored via sensitivity analyses,
meta- regression and subgroup analyses. The robustness of the
results was examined by excluding those rated as at a higher risk
of bias. Publication bias was assessed following two steps: (1)
testing the symmetry of the funnel plot by the Egger method22;
and (2) determining whether any asymmetry was due to publi-
cation bias via enhanced- contour funnel plots after the trim-
and- fill method.23 The meta- regression and publication bias test
were conducted only when at least 10 studies were available.
The “metagen”, “metabias” and “trimfill” packages in R 3.4.3
software (https://www. r- project. org) were used to perform all
the analyses.
Additionally, multiple subgroup and sensitivity analyses were
conducted to take into account the following cases where results
might be biased. First, 82% of studies recruited people without
dementia at baseline and only 17% specifically constrained the
population to those with normal cognition. Notably, inclusion of
individuals with mild cognitive impairment, who might be at a
prodromal stage of AD, resulted in a degree of misclassification
bias, especially when the population was at an advanced age and
was insufficiently followed. Thus, subgroup analyses according
to the cognitive status at baseline (free of dementia vs cogni-
tively normal), sufficiency of follow- up (online supplementary
appendix 1) and life stage were performed. Second, it was often
clinically difficult to distinguish mixed AD (coexistence of AD
and vascular dementia (VD)) from VD among elderly people,
especially when the pathological evidence is often unavailable
and the individual has a history of stroke. Thus, to examine the
influence of potential misclassification bias, subgroup analyses
based on AD outcomes (all AD vs probable or pure AD (p- AD)
defined as AD without VD or cerebrovascular disease (CVD))
were performed. Third, sensitivity analyses excluding studies
with high attrition rates and poor generalisability (online supple-
mentary appendix 1) were conducted.
Patient involvement
No patients were involved in setting the research question or the
outcome measures, nor were they involved in developing plans
for design or implementation of the study. No patients were
asked to advise on interpretation or writing up of results. There
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from
4Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
Cognitive neurology
Figure 3 Risk of bias profile, meta- analysis results, sample size (figure3A), credibility rating (figure3B) and summary (figure3C) for 43 significant
modifiable risk factors based on observational prospective studies. When the mean score (for each bias domain) ≤0.5 was regarded as possibly moderate-
to- high risk, analyses for 79% of factors had problems of generalisability, 60% for high attrition, 48% for insufficient follow- up, 40% for reverse causality,
8% for confounding bias and 6% for assessment of exposure. For a summary of the effect, a total of 43 factors showed significant associations with AD risk;
26 risk factors and eight protective factors were identified that modify the risk by at least 25% (figure3A). For credibility of the pooled results, 11 factors
were rated at a moderate- to- high level (G, G/A+ or A+ level), 20 were rated at a low- to- moderate level (A+/A− or A− level) and 12 were rated at a very low
level (S+, S− or P level) (figure3B). With good performance in all the domains above, eight risk factors are highlighted (figure3C). AD, Alzheimer’s disease;
BMD, bone mineral density; BMI, body mass index; CVD, cerebrovascular disease; DBP, diastolic blood pressure; HSV, herpes simplex virus; IMT, intima- media
thickness; NSAIDs, non- steroidal anti- inflammatory drugs; SHBG, sex hormone binding globulin.
are no plans to disseminate the results of the research to study
participants or the relevant patient community. No evaluation
was undertaken to determine whether the studies included in the
review had any patient involvement.
RESULTS
Figure 1 shows the flow diagrams of the study selection process
for OPSs (figure 1A) and RCTs (figure 1B). The search yielded
33 145 and 11 531 records for OPSs and RCTs, respectively.
After integration with the AlzRisk database and Clinicaltrials.
gov website, a total of 243 OPSs and 153 completed RCTs
were finally included. Evidence- based profiles were constructed
(online supplementary appendix 3 & 4). The global distribution
of studies eligible for the systematic review and their character-
istics are shown in figure 1C. The sources of bias for the current
evidence profile mainly consisted of generalisability, attrition
and misclassification for OPSs and performance bias, incomplete
outcome data, inadequate allocation concealment and selective
outcome reporting for RCTs (online supplementary appendix
figure 1).
Meta- analyses were conducted for 134 risk factors (online
supplementary appendix 5). A total of 43 factors showed signif-
icant associations with the risk of AD, among which 80% were
identified as significantly modifying the risk by at least 25%
(figure 3A). Indicating the credibility of pooled results, anal-
yses for eight risk factors (diabetes, orthostatic hypotension,
hypertension in midlife, head trauma, stress, depression, midlife
obesity and coronary artery bypass grafting (CABG) surgery) and
three protective factors (cognitive activity, increased BMI in late
life and education) were rated with moderate- to- high level cred-
ibility (G, G/A+ or A+ level). In addition, 20 factors were rated
at a low- to- moderate level (A+/A− or A− level) and 12 were
rated at a very low level (S+, S− or P level) (figure 3B). With
good performance in all the domains above, eight factors were
highlighted, including depression (A+ level; RR 1.80; 95% CI
1.34 to 2.42), CABG surgery (G/A+ level; RR 1.71; 95% CI
1.04 to 2.79), diabetes mellitus (G level; RR 1.69; 95% CI 1.51
to 1.89), stress (G/A+ level; RR 1.56; 95% CI 1.19 to 2.04),
hypertension in midlife (G/A+ level; RR 1.38; 95% CI 1.29 to
1.47), head trauma (G/A+ level; RR 1.35; 95% CI 1.18 to 1.54),
cognitive activity (A+ level; RR 0.50; 95% CI 0.39 to 0.63) and
more formal schooling years (>6 to 15 years) (G level; RR 0.49;
95% CI 0.40 to 0.62) (figure 3C). Additionally, another 91 items
were found to impart no influence on the risk of AD, but mostly
with low levels of credibility, except for late- life hypertension
(G level, RR 0.96; 95% CI 0.79 to 1.17) (online supplementary
appendix figure 2).
For RCTs, 29 meta- analyses covering 11 interventions were
conducted (online supplementary appendix 6). Three interven-
tions, including total homocysteine (tHcy)- lowering treatment
(using folic acid, vitamin B12 and vitamin B6), cocoa flavanol
and physical activity showed significant associations with AD or
cognitive endpoints. For the directness of the outcomes, only
five meta- analyses (involving acetylcholinesterase inhibitor, anti-
hypertensive treatment, non- steroidal anti- inflammatory drugs
(NSAIDs), hormone replacement therapy and ginkgo biloba)
examined associations with AD (figure 4A). For the levels of cred-
ibility, nine meta- analyses were rated at a moderate- to- high level
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from
5
Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
Cognitive neurology
Figure 4 Risk of bias profile, meta- analysis results, sample size (figure4A), credibility rating (figure4B) and summary (figure4C) for 11 interventions
based on randomised controlled trials. When the mean score (for each bias domain) ≤0.5 was regarded as possibly moderate- to- high risk, 17.2% meta-
analyses had problems of inadequate concealment of allocations, 27.6% for performance bias, 3.4% for detection bias, 24.1% for incomplete outcome
data, 13.8% for selective outcome reporting and 31% for other sources of bias. For the significance of the pooled results, six meta- analyses showed
significant associations (figure4A). For credibility of the pooled results, nine meta- analyses were rated at a moderate- to- high level (G, G/A+ or A+ level),
three at a low- to- moderate level (A+/A− or A− level) and 17 at a very low level (S+, S− or P level). Specifically, moderate- to- high credibility of results
showed little benefit on the risk of Alzheimer's disease from acetycholinesterase inhibitors, antihypertensive agents in late life, oestrogen therapy, and
DHA+EPA supplementation. No robust conclusion could be reached for non- steroidal anti- inflammatory drugs, ginkgo biloba, cocoa flavanol and cognitive
training. For directness of outcomes, five meta- analyses examined the associations with AD (figure4B). Although none showed a good performance in all the
above domains, two interventions (physical exercise and total homocysteine- lowering treatment) seem more promising than others (figure4C).
(G, G/A+ or A+ level), three were rated at a low- to- moderate
level (A+/A− or A− level) and 17 were rated at a very low level
(S+, S− or P level) (figure 4B). The overall evaluation high-
lighted two interventions that seemed promising (figure 4C):
physical exercise (mini- mental state examination (MMSE), stan-
dardised mean difference (SMD) 0.28, 95% CI 0.07 to 0.50 and
AD assessment scale cognition, SMD 0.25, 95% CI 0.08 to 0.41)
and tHcy- lowering treatment (MMSE, SMD 0.09, 95% CI 0.02
to 0.15) (online supplementary appendix figure 3). Notably,
oestrogen therapy was associated with an increase in the risk of
dementia (G level).
The significance and the effect size minimally changed for most
factors after excluding ORs (online supplementary appendix
figure 4). No influences of publication bias on the pooled
results were identified (online supplementary appendix 5). The
sources of heterogeneity were explored. For diabetes (n=14,
I2=65%), the percentage of women explained 39% heteroge-
neity (p=0.008), which might be attributed to inclusion of two
high- risk- of- bias studies24 25 that explored associations only for
men. The mean age at baseline explained most heterogeneity for
hypertension (p=0.0003) and BMI (p=0.091, τ2=0). No influ-
ences of lowering the heterogeneity (I2 <10%) via sensitivity
analyses on the pooled results were found for current smoking,
systolic blood pressure, education and depression. The influence
of risk of bias might be low for depression while smoking and
stroke were vulnerable to sources of bias due to misclassification,
attrition and generalisability (online supplementary appendix
figure 5).
Twenty- one evidence- based suggestions with different levels
of evidence (11 with Level A and 10 with Level B) and strength
of suggestions (19 with Class I and two with Class III) are listed
in table 1. Specifically, Class I suggestions were for 19 factors,
including 10 factors with Level A evidence (cognitive activity,
hyperhomocysteinaemia, increased BMI in late life, depression,
stress, diabetes, head trauma, hypertension in midlife, ortho-
static hypotension and education) and nine factors with Level
B evidence (obesity in midlife, weight loss in late life, physical
exercise, smoking, sleep, CVD, frailty, atrial fibrillation and
vitamin C) (figure 5). Two factors were not recommended (Class
III): oestrogen replacement therapy (Level A) and acetylcholin-
esterase inhibitors (Level B) (online supplementary appendix
7 & appendix figure 6). Six factors (diastolic blood pressure
management, NSAID use, social activity, osteoporosis, pesticide
exposure and silicon from drinking water) were rated as Level
C low- strength evidence, with the recommendation that their
relationships with AD be confirmed in future studies.
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from
6Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
Cognitive neurology
Table 1 Guideline for prevention of AD: preliminary clinical suggestions*
Factors/interventions Suggestion
Lifestyle
BMI and weight
management
►Adults aged <65 years should maintain or lose weight through an appropriate balance of physical activity, caloric intake and formal behavioural programmes
when indicated to maintain/achieve a BMI between 18.5 and 24.9 kg/m2 (Class I, level B)
►Adults aged >65 years should not to be too skinny (Class I, level A4)
►Adults aged >65 years with a trend of weight loss should be closely monitored for their cognitive status (Class I, Level B)
Physical exercise ►Individuals, especially those aged ≥65 years, should stick to regular physical exercise (Class I, Level B*)
Cognitive activity ►Mentally stimulating activities should be encouraged, such as reading, playing chess, etc (Class I, Level A4)
Smoking ►People should not smoke and should avoid environmental tobacco smoke. Ccounselling, nicotine replacement and other pharmacotherapy as indicated
should be provided in conjunction with a behavioural programme or formal smoking cessation programme (Class I, Level B)
Sleep ►Get sufficient and good quality sleep and consult a doctor or receive treatment when you have problem with sleep (Class I, Level B)
Comorbidities
Diabetes ►Stay away from diabetes via a healthier lifestyle and diabetic patients should be closely monitored for their cognitive decline (Class I, Level A4)
CVD ►Maintain a good condition of the cerebral vessels via a healthier lifestyle or medications to avoid atherosclerosis, low cerebral perfusion and any CVD.
Individuals with stroke, especially cerebral microbleeding, should be carefully monitored for their cognitive change and take preventative measures as
indicated to protect cognition (Class I, level B)
Head trauma ►Protect your head from injuries (Class I, level A4)
Frailty ►Stay healthy and strong in late life. Those with increasing frailty should be especially monitored for their cognition (Class I, Level B)
Blood pressure ►Individuals aged < 65 years should avoid hypertension via a healthier lifestyle (Class I, Level A4)
►Individuals with OH should be closely monitored for their cognition (Class I, Level A4)
Depression ►Maintain a good condition of mental health and closely keep an eye on the cognitive status for those with depressive symptoms (Class I, Level A4)
AF ►Maintain a good cardiovascular condition and manage AF using pharmaceuticals (Class I, level B)
Stress ►Relax your mind and avoid daily stress (Class I, Level A4)
Other domains
Education ►Receive as much education as possible in early life (Class I, level A4)
Hyperhomocysteinaemia ►Have a regular blood examination for homocysteine level. Individuals with hyperhomocysteinaemia should be treated with vitamin B and/or folic acid and be
followed with a focus on their cognition (Class I, Level A2)
Vitamin C ►Vitamin C in the diet or taken as supplements might help (Class I, Level B)
Not recommended
ERT ►Oestrogen replacement therapy should not be specifically used for AD prevention in postmenopausal women (Class III, Level A2)
ACI ►ACI should not be used for AD prevention in cognitively impaired individuals (Class III, Level B)
*The risk of bias is rated as high mainly due to lack of a blinding method and allocation concealment, which however cannot be achieved in randomised controlled trials for interventions such as
physical exercise. We therefore consider that the results are relatively more reliable than rated. Also, the content cannot be too detailed (especially for the dose and duration) for some factors and
a very good trial is needed to replicate (pivotal studies). Also, these suggestions must be presented in the context of the limitations of the studies and continuing uncertainty among investigators.
ACI, acetylcholinesterase inhibitors; AF, atrial fibrillation; BMI, body mass index; CVD, cerebrovascular disease; ERT, estrogen replacement therapy; IMT, intima- media thickness; NSAIDs, non-
steroidal anti- inflammatory drugs; OH, orthostatic hypotension.
DISCUSSION
Our systematic review and meta- analysis identified a total of 21
evidence- based suggestions that can be used in life- course prac-
tices to prevent AD. Nineteen were regarded as ‘strong sugges-
tions’, nine of which were rated with Level A evidence (table 1).
Nearly two- thirds of these suggestions target vascular risk
factors and lifestyle, strengthening the importance of keeping a
good vascular condition and maintaining a healthy lifestyle for
preventing AD.
Strengths and weaknesses of this study
This is the most comprehensive and large- scale systematic review
and meta- analysis for AD prevention to date. The evidence-
based suggestions are constructed by integrating a large amount
of evidence from both OPSs and RCTs. Sources of bias and
robustness of evidence were thoroughly assessed and secondary
analyses were used to explore their influences, guaranteeing the
objectivity and transparency of our findings. Furthermore, the
outcome of OPSs was confined to AD dementia, given that the
heterogeneity of endpoints might complicate the profile and
downgrade the credibility of the evidence because: (1) observa-
tional studies are more vulnerable to sources of bias than RCTs,
even though a rigorous procedure was employed to grade the
evidence; (2) non- AD dementia accounts for roughly 30% of
incident dementia (online supplementary appendix figure 7) and
the false positive rate for diagnosis of mild cognitive impairment
is fairly common.26 Some caveats should also be emphasised.
Observational studies cannot indicate causal relationships and
RCTs may not be generalisable beyond the specific sample, inter-
vention, dose and duration studied. Classification of the avail-
able evidence including assessment of potential biases requires
subjective judgement. The values of the current suggestions
might be confined by geographic variability, definition of expo-
sure and prevalence of risk factors at the population level. Some
important factors of all- cause dementia were inadequately inves-
tigated for AD, such as social determinants27 and frailty,28 and
more high- quality prospective studies are warranted to bridge
this gap. AD is challenging to study. The neurobiology of AD
begins at least 15 years before symptoms appear. Tools such as
amyloid and tau PET scanning are available to characterise the
neuropathology at any stage, but it is impractical to include such
assessments in large observational studies; without biomarker
data, misclassification is unavoidable and several conclusions
may be challenged by studies in the near future. Despite these
challenges, this systematic review and meta- analysis can suggest
recommendations to guide clinicians, even as the field perseveres
with additional studies. These evidence- based suggestions must
be presented in the context of the limitations of the studies and
continuing uncertainty among investigators. Finally, the present
study did not register and the protocol can be found in online
supplementary appendix 8.
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from
7
Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
Cognitive neurology
Figure 5 Distribution of modifiable factors with Class I recommendation throughout the course of life. Class I suggestions (benefit >>risk due to
intervention) risk factors include 10 factors with Level A evidence (cognitive activity, hyperhomocysteinaemia, increased BMI in late life, depression, stress,
diabetes, head trauma, hypertension in midlife, orthostatic hypotension and education) and 9 factors (obesity in midlife, weight loss in late life, physical
exercise, smoking, sleep, CVD, frailty, atrial fibrillation and vitamin C) with Level B evidence. The x axis represents the mean age of the total sample (solid
circle) with a range of mean age (short horizontal line) for observational prospective studies included. The y axis represents the summary relative risk (RR).
AD, Alzheimer’s disease; OH, orthostatic hypotension; CVD, cerebrovascular disease; IMT, intima- media thickness.
Strengths and weaknesses in relation to other studies
Notably, tHcy- lowering treatment seems the most promising
intervention for AD prevention, in agreement with a recent
report.29 The Lancet Commission on dementia has recently
proposed nine potentially modifiable risk factors of all- cause
dementia. However, these suggestions might not be directly
applicable to AD, bearing in mind that the heterogeneity of
endpoints complicates the profile and reduces the credibility
of evidence for AD prevention. Our study generated more
evidence- based suggestions associated with a decreased risk of
AD, filling this gap in the field.
Meaning of the study
The hypotheses for the underlying mechanisms may include
brain reserve theory, the hypoperfusion hypothesis, one- carbon
methabolism, hypomethylation theory, inflammation and the
oxidative stress hypothesis. The combination of multiple recom-
mendations is most likely the best approach to delay the onset
of AD, as indicated by the Finnish Geriatric Intervention Study
to Prevent Cognitive Impairment and Disability (FINGER).30
On the basis of this paper, future clinical trials should focus
on exploring the best combination of recommendations with
Class I recommendation and Level A evidence to prevent AD
using larger samples, particularly in real- world settings. These
evidence- based suggestions should be particularly noted by
non- demented but high- risk individuals (eg, people with
AOPEε4, a high polygenic score, a family history of dementia or
amyloid- positive evidence31) and family doctors to give optimal
recommendations to their patients in terms of what they might
do to get the best protection against AD.
Future research
For OPSs, low participation rates (cognitive activity and stroke),
high attrition (stroke, smoking, alcohol drinking and hyperten-
sion) and follow- up insufficiency (stroke and smoking) should
be specifically highlighted in future prospective studies. Reverse
causality might bias the association with late life obesity.32 It
is unclear whether reverse causality exists for other potential
factors such as frailty, social isolation and sleep disorders. Investi-
gation and comparison of important characteristics of those who
refused to participate or were lost during follow- up might be a
good method to guarantee optimised validity. Subgroup effects
exist due to the characteristics of the sample (eg, age, gender,33
APOEε4 status34 and medication compliance34) or exposure (eg,
type, dose and duration). For RCTs, choosing the suitable popu-
lation might be the key to determining whether an intervention
can work. The optimal time window also matters,35 especially
considering that benefits were weak for those with a clinical diag-
nosis of dementia.36 Generalisablity should be further optimised,
such as recruiting larger samples from community- dwelling indi-
viduals and searching for methods to lower dropout rates. Well-
designed clinical trials are needed to verify the effects on AD of
several promising interventions, including sleep improvement,
smoking cessation, antidepression management and antidiabetic
agents.
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from
8Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
Cognitive neurology
CONCLUSIONS
Twenty- one clinical evidence- based suggestions are proposed,
offering clinicians and stakeholders an evidence- based guideline
for AD prevention. With credible though inconclusive evidence,
the suggestions targeted 10 risk factors including diabetes,
hyperhomocysteinaemia, poor BMI management, reduced
education, hypertension in midlife, orthostatic hypotension,
head trauma, less cognitive activity, stress and depression. This
study provides an advanced and contemporary survey of the
evidence, suggesting that more high- quality OPSs and RCTs are
urgently needed to strengthen the evidence base for uncovering
more promising approaches to preventing AD.
Author affiliations
1Department of Neurology and Institute of Neurology, Huashan Hospital, Shanghai
Medical College, Fudan University, Shanghai, China
2Department of Neurology, Qingdao Municipal Hospital, Qingdao University,
Qingdao, China
3Department of Epidemiology and Public Health, University of Toulouse III, Toulouse,
France
4Department of Psychiatry, Medical Research Council and Wellcome Trust
Behavioural and Clinical Neuroscience Institute, University of Cambridge, Cambridge,
UK
5Department of Hygiene and Epidemiology, University of Ioannina Medical School,
Ioannina, Greece
6Department of Epidemiology and Biostatistics, School of Public Health, Tongji
Medical College, Huazhong University of Science and Technology, Wuhan, China
7Department of Neurology, Massachusetts General Hospital and Harvard Medical
School, Charlestown, Massachusetts, USA
8Department of Neurology, Xuan Wu Hospital, Capital Medical University, Beijing,
China
9Department of Psychological Medicine, Yong Loo Lin School of Medicine, National
University of Singapore, Singapore
10Department of Neurology, Daping Hospital, Third Military Medical University,
Chongqing, China
11Department of Medicine and Therapeutics, Faculty of Medicine, The Chinese
University of Hong Kong, Shatin, Hong Kong
12Department of Neurology, University Hospital of Montpellier, Montpellier, France
13McGill Center for Studies in Aging, McGill University, Montreal, Quebec, Canada
14Alzheimer’s Therapeutic Research Institute, University of Southern California, San
Diego, California, USA
15Department of Geriatrics, Purpan University Hospital, Toulouse, France
Twitter Wei Xu @na and Can Zhang @martindragoncn
Acknowledgements The authors thank Professor Michael M Weiner and Dr Yu-
Tzu Wu for critical review.
Contributors JTY and BV conceived and designed the study. JTY, WX, CCT, HFW,
MST and JQL selected the articles and extracted the data. JTY, WX, C- CT, HFW, M- ST,
J- QL, XHH, YW, LT and LT proofread the data. JTY, WX, C- CT and AP analysed the
data. WX and CCT generated the figures. JTY, WX, CCT wrote the first draft of the
manuscript. JTY, WX, C- CT, SA, JS, EE, CZ, JJ, AP, LF, EHK, YJW, VM, JT, GS, PSA, QD,
and BV interpreted the data and contributed to the writing of the final version of the
manuscript. All authors agreed with the results and conclusions and approved the
final draft.
Funding This study was supported by grants from the National Key R&D Program
of China (2018YFC1314702), Shanghai Municipal Science and Technology Major
Project (No. 2018SHZDZX03) and ZHANGJIANG LAB.
Competing interests JTY serves as an associate editor- in- chief for Annals of
Translational Medicineand is senior editor for Journal of Alzheimer’s Disease. SA has
received grants from Europe, Ipsen, and France Alzheimer, served as a consultant
for Ipsen, Pierre Fabre, Lilly, Nestlé, Sanofi and Servier, and received non- financial
support from Biogen, Nutrition Santé, Pfzer and Icon, and other support from
the AMPA Association. GS has received clinical trial support from Lilly and Roche
in DIAN- TU, TauRx Therapeutics (TauRx) and Lundbeck; has been a data safety
monitoring board (DSMB) member of ADCS, ATRI, API and Eisai; and has been
a scientific adviser to Affiris, Boehringer Ingelheim, Lilly, Roche, Servier, Sanofi,
Schwabe, Takeda and TauRx. PSA has received grants from the US Alzheimer’s
Association, Janssen, Lilly, the US National Institute on Aging and Toyama; and
consulting fees from Abbott, Abbvie, Amgen, Anavex, AstraZeneca, Biogen Idec,
Biotie, Bristol- Myers Squibb, Cardeus, Cohbar, Eisai, Elan, Eli Lilly, Genentech,
Ichor, iPerian, Janssen, Lundbeck, Medivation, Merck, NeuroPhage, Novartis, Pfizer,
Probiodrug, Roche, Somaxon and Toyama, outside the submitted work. BV reports
grants from Pierre Fabre, Avid, Exonhit, AbbVie, Lilly, Lundbeck, MSD, Otsuka,
Regenron, Sanofi, Roche, AstraZeneca, LPG Systems, Nestlé and Alzheon, and
personal fees from Lilly, Lundbeck, MSD, Otsuka, Roche, Sanofi, Biogen, Nestlé,
Transition Therapeutics and Takeda.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Open access This is an open access article distributed in accordance with the
Creative Commons Attribution Non Commercial (CC BY- NC 4.0) license, which
permits others to distribute, remix, adapt, build upon this work non- commercially,
and license their derivative works on different terms, provided the original work is
properly cited, appropriate credit is given, any changes made indicated, and the use
is non- commercial. See:http:// creativecommons. org/ licenses/ by- nc/ 4. 0/.
ORCID iDs
Jin- TaiYu http:// orcid. org/ 0000- 0002- 7686- 0547
WeiXu http:// orcid. org/ 0000- 0002- 3310- 5875
RefeRences
1 Qiu C, von Strauss E, Bäckman L, etal. Twenty- year changes in dementia
occurrence suggest decreasing incidence in central Stockholm, Sweden. Neurology
2013;80:1888–94.
2 Matthews FE, Arthur A, Barnes LE, etal. 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
2013;382:1405–12.
3 Satizabal CL, Beiser AS, Chouraki V, etal. Incidence of dementia over three decades in
the Framingham Heart Study. N Engl J Med 2016;374:523–32.
4 Wu Y- T, Fratiglioni L, Matthews FE, etal. Dementia in Western Europe: epidemiological
evidence and implications for policy making. Lancet Neurol 2016;15:116–24.
5 Andrieu S, Coley N, Lovestone S, etal. Prevention of sporadic Alzheimer’s
disease: lessons learned from clinical trials and future directions. Lancet Neurol
2015;14:926–44.
6 Livingston G, Sommerlad A, Orgeta V, etal. Dementia prevention, intervention, and
care. Lancet 2017;390:2673–734.
7 Winblad B, Amouyel P, Andrieu S, etal. Defeating Alzheimer’s disease and
other dementias: a priority for European science and society. Lancet Neurol
2016;15:455–532.
8 Mazure CM, Swendsen J. Sex differences in Alzheimer’s disease and other dementias.
Lancet Neurol 2016;15:451–2.
9 Coley N, Andrieu S, Gardette V, etal. Dementia prevention: methodological
explanations for inconsistent results. Epidemiol Rev 2008;30:35–66.
10 Moher D, Liberati A, Tetzlaff J, etal. Preferred reporting items for systematic reviews
and meta- analyses: the PRISMA statement. Int J Surg 2010;8:336–41.
11 Stroup DF, Berlin JA, Morton SC, etal. Meta- analysis of observational studies in
epidemiology: a proposal for reporting. Meta- analysis Of Observational Studies in
Epidemiology (MOOSE) group. JAMA 2000;283:2008–12.
12 Weuve JMM, Blacker D. The AlzRisk database. Alzheimer research forum. Available:
http://www. alzforum. org [Accessed [date of access].
13 McKhann G, Drachman D, Folstein M, etal. 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.
14 . Available: http:// community. cochrane. org/ editorial- and- publishing- policy- resource/
cochrane- review- development/ cochrane- handbooks/ cochrane- handbook- systematic-
reviews- interventions
15 Deeks JJ, Dinnes J, D’Amico R, etal. Evaluating non- randomised intervention studies.
Health Technol Assess 2003;7:27.
16 Guyatt GH, Oxman AD, Vist G, etal. GRADE guidelines: 4. Rating the quality of
evidence--study limitations (risk of bias). J Clin Epidemiol 2011;64:407–15.
17 Guyatt GH, Oxman AD, Kunz R, etal. GRADE guidelines: 7. Rating the quality of
evidence--inconsistency. J Clin Epidemiol 2011;64:1294–302.
18 Guyatt GH, Oxman AD, Kunz R, etal. GRADE guidelines 6. Rating the quality of
evidence--imprecision. J Clin Epidemiol 2011;64:1283–93.
19 Higgins JPT, Thompson SG, Deeks JJ, etal. Measuring inconsistency in meta- analyses.
BMJ 2003;327:557–60.
20 Riley RD, Higgins JPT, Deeks JJ. Interpretation of random effects meta- analyses. BMJ
2011;342:d549.
21 Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta- analysis. Stat Med
2002;21:1539–58.
22 Egger M, Davey Smith G, Schneider M, etal. Bias in meta- analysis detected by a
simple, graphical test. BMJ 1997;315:629–34.
23 Peters JL, Sutton AJ, Jones DR, etal. Contour- enhanced meta- analysis funnel plots
help distinguish publication bias from other causes of asymmetry. J Clin Epidemiol
2008;61:991–6.
24 Moffat SD, Zonderman AB, Metter EJ, etal. Free testosterone and risk for Alzheimer
disease in older men. Neurology 2004;62:188–93.
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from
9
Yu J- T, etal. J Neurol Neurosurg Psychiatry 2020;0:1–9. doi:10.1136/jnnp-2019-321913
Cognitive neurology
25 Dal Forno G, Palermo MT, Donohue JE, etal. Depressive symptoms, sex, and risk for
Alzheimer’s disease. Ann Neurol 2005;57:381–7.
26 Koepsell TD, Monsell SE. Reversion from mild cognitive impairment to normal or near-
normal cognition: risk factors and prognosis. Neurology 2012;79:1591–8.
27 Fratiglioni L, Wang HX, Ericsson K, etal. Influence of social network on occurrence of
dementia: a community- based longitudinal study. Lancet 2000;355:1315–9.
28 Stower H. Linking Alzheimer’s disease and frailty. Nat Med 2019;25:358.
29 Kane RL, Butler M, Fink HA, etal. Interventions to prevent age- related cognitive
decline, mild cognitive impairment, and clinical Alzheimer's- type dementia. Rockville
(MD, 2017.
30 Ngandu T, Lehtisalo J, Solomon A, etal. 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 2015;385:2255–63.
31 Donohue MC, Sperling RA, Petersen R, etal. Association between elevated brain
amyloid and subsequent cognitive decline among cognitively normal persons. JAMA
2017;317:2305–16.
32 Kivimäki M, Luukkonen R, Batty GD, etal. Body mass index and risk of dementia:
analysis of individual- level data from 1.3 million individuals. Alzheimers Dement
2018;14:601–9.
33 Zissimopoulos JM, Barthold D, Brinton RD, etal. Sex and race differences in the
association between statin use and the incidence of Alzheimer disease. JAMA Neurol
2017;74:225–32.
34 Rajan KB, Barnes LL, Wilson RS, etal. BP and risk of incident AD dementia by
antihypertensive and APOE epsilon4 allele. Ann Neurol 2018;83:935–44.
35 Kryscio RJ. Secondary prevention trials in Alzheimer disease: the challenge of
identifying a meaningful end point. JAMA Neurol 2014;71:947–9.
36 Lamb SE, Sheehan B, Atherton N, etal. Dementia and physical activity (DAPA) trial of
moderate to high intensity exercise training for people with dementia: randomised
controlled trial. BMJ 2018;361:k1675.
on July 21, 2020 by guest. Protected by copyright.http://jnnp.bmj.com/J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2019-321913 on 20 July 2020. Downloaded from