is high homocysteine level a risk factor for cognitive decline in elderly? a systematic review,...

REGULAR RESEARCH ARTICLES

Is High Homocysteine Level a Risk Factorfor Cognitive Decline in Elderly? A

Systematic Review, Meta-Analysis, andMeta-Regression

Roger C. M. Ho, D.P.M., M.R.C. Psych., Mike W. L. Cheung, Ph.D.,Erin Fu, B. Soc. Sc., Hlaing H. Win, M.B.B. S., Min Htet Zaw, M.B.B.S.,

Amanda Ng, B.Sc., Anselm Mak, M.Med.Sc., F.R.CP

Obective: High homocysteine (Hct) has been causatively linked to Alzheimer disease(AD) and vascular dementia (VaD) in old age, but research methodologies and out-come measures are heterogeneous. It remains unclear whether the findings can be gen-eralized across studies. Methods: Random-effects meta-analyses were conducted onstudies examining the relationship between Hct level and risk of developing dementia/cognitive decline between comparison groups. Meta-regression identified patient- andtrial-related factors, which may contribute to heterogeneity. Results: Seventeen rele-vant studies (6,122 participants; 13 cross-sectional and fourprospective studies) wereincluded. Compared with controls, Hct was significantly elevated in AD (pooled stan-dardized mean difference [SMD]: 0.59; 95% confidence interval [CI]: 0.38–0.80; sig-nificant heterogeneity: τ2 = 0.105) and VaD (pooled SMD: 1.30; 95% CI: 0.75–1.84;significant heterogeneity: τ2 = 0.378). Meta-regression identified mean age as sig-nificant moderator for AD versus controls and mean age and mean folate levelsas significant moderators for VaD versus controls. Hct was significantly higher inVaD relative to AD (pooled SMD: 0.48; 95% CI: 0.23–0.73; moderately significant het-erogeneity: τ2 = 0.076); proportion of men and mean folate levels were significantmoderators. High-Hct level was not associated with risk of developing dementia inprospective studies (pooled odds ratio: 1.34; 95% CI: 0.94–1.91, nonsignificant het-erogeneity: τ2 = 0.048). Conclusion: Individuals with AD and VaD have higher Hctlevels than controls; however, a causal relationship between high-Hct level and riskof developing dementia is not supported. More prospective studies and randomizedcontrolled trials are required to test the therapeutic benefits of lowering Hct levels.(Am J Geriatr Psychiatry 2011; 19:607–617)

Key Words: Homocysteine, dementia, elderly, meta-analysis, systematic review

Received October 22, 2009; revised May 3, 2010; accepted June 3, 2010. From the Department of Psychological Medicine, Yong Loo Lin Schoolof Medicine (RCMH, EF, HHW, AN); Department of Psychology, Faculty of Arts and Social Science (MWLC); Department of Cardiac, Thoracicand Vascular Surgery, Yong Loo Lin School of Medicine (MHZ); and Department of Medicine, Yong Loo Lin School of Medicine (AM), NationalUniversity of Singapore, Singapore. Send correspondence and reprint requests to Dr. Roger C.M. Ho, Department of Psychological Medicine,National University Hospital, 5 Lower Kent Ridge Road, Singapore 119074. e-mail: [email protected]

c© 2011 American Association for Geriatric PsychiatryDOI: 10.1097/JGP.0b013e3181f17eed

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Am J Geriatr Psychiatry 19:7, July 2011 607

Is High Homocysteine Level

D ementia is one of the most disabling and bur-densome health conditions worldwide.1 It has

been estimated that 24 million people currently suf-fer from dementia and that the number of patientswill double every 20 years to 81 million by the Year2040. Prevention of dementia at the population levelrequires the identification of modifiable risk factors.Homocysteine (Hct) was discovered by Butz and duVigneaud in 1932, and the origin of its name camefrom its chemical structure as it is the next highersymmetrical homologue of cystine.2

High-Hct levels have been hypothesized to increasethe risk of dementia. The evidence linking high-Hctstatus and dementia seems to come from two comple-mentary sources of evidence. First, a significant per-centage of patients with dementia have been reportedto have high-Hct levels. Second, links between in-creased Hct and white matter lesions, oxidativestress, and amyloid formulation have been postu-lated with the advent of neurobiological research.2

It is also unclear whether the association with Hct issimilar for dementias with different etiologies.

The existence of a possible relationship betweenhigh-Hct status and dementia has been examined indetail using observational epidemiological designs.Reviews on the relationship between Hct and cog-nitive impairment have been published,2,3 but therehas been no published meta-analysis on this topic.Reduced appetite—an occasional feature of demen-tia, unhealthy dietary patterns,4 nutritional factors,5

baseline functional impairment,6 and deficiency in in-trinsic factors—may lead to impaired nutritional sta-tus deficient in Vitamin B-12 and folate and subse-quently confound any relationship between Hct anddementia. In addressing these issues, we performeda systematic review and meta-analysis to crystallizecurrent knowledge concerning the role of Hct as arisk factor for dementia, cognitive decline, and cog-nitive impairment. Meta-regression was used to ex-amine sources of heterogeneity between studies,7 in-cluding the effects of moderators such as the levels ofVitamin B-12 and folate on observed associations.

METHODS

Search Strategy

The following online databases were searchedfrom inception to December 2008: Pubmed, Embase,

PsychINFO, BIOSIS, and Science Direct andCochrane CENTRAL. The search terms used were“Homocysteine” and a combination of keywords forcognition: “cognit∗, memory, attention, reaction time,speed of processing, crystallized ability, crystallizedintelligence, fluid ability, fluid intelligence, generalmental ability, intelligence, executive function, neu-ropsychological testing, mini mental state examina-tion, dementia, mild cognitive impairment (MCI),Alzheimer disease” (where ∗ indicates truncation).8

The search was limited to English articles and articlesthat reported data from humans. Abstracts presentedin major international conferences were manuallysearched, and the authors of correspondence werecontacted for further information.

Inclusion Criteria

We sought all cross-sectional, case–control, andprospective cohort epidemiological designs, whichexamined the relationship between Hct level and de-mentia and cognitive decline.

Studies were included if they 1) were case–controlstudies or were prospective with at least a 12-monthfollow-up period; 2) analyzed a blood measure ofHct as the main variable of interest or as a covariate;3) analyzed dementia or cognitive decline as thedependent variable where cognition or dementiawere assessed clinically (i.e., subjective assessmentsof cognition were not acceptable for inclusion norwere dementia diagnoses obtained from medicalrecords); and 4) had sufficient information to calcu-late effect size.

The cognitive outcomes included a) cognitive per-formance at follow-up; (b) cognitive performancechange, which was the difference in cognitive perfor-mance between baseline and follow-up; and (c) cog-nitive decline where the sample was dichotomizedinto dementia and nondementia groups. In the sam-ple group with dementia, two potential dementiaoutcomes were addressed, including Alzheimer dis-ease (AD) and vascular dementia (VaD).

For case-control studies to be included, studiesmust have drawn some comparison between thepresence of high-Hct levels and dementia status andmust have used some comparison or control group.For inclusion of prospective studies in the meta-analysis, sufficient information was required to esti-mate odds ratio (OR).


608 Am J Geriatr Psychiatry 19:7, July 2011

Ho et al.

Exclusion Criteria

Studies were excluded if the sample mean age was<60 years or if the authors only reported the corre-lations between Hct levels and cognitive functionswithout reporting the mean Hct levels in the demen-tia and nondementia groups.

We conducted a systematic review and meta-analysis in line with best practice guidelines on thesynthesis of observational epidemiological data.9,10

Selection of relevant publications was conducted in-dependently by two authors, and any disagreementswere resolved through discussions. Articles were dei-dentified (blinded title, author(s), year of publication,and journal name) before data extraction. The follow-ing information was extracted from each article andcrosschecked by the second researcher: average ageof participants, proportion of men, average years ofeducation, smoking status, length of follow-up, de-scription of Hct measure, description of cognitive ordementia measures, and the relevant statistics de-scribing the relationship between Hct and cognitionor dementia. Summary statistics, such as group meanHct levels and OR, with 95% confidence intervals(CIs), were extracted or calculated from the originaldata.

Statistical Analysis

All statistical analyses were performed with Com-prehensive Meta-Analysis Version 2.0, developed forsupport in meta-analysis. The figures were preparedwith the Metafor package implemented in the Renvironment.11,12 Differences in mean Hct levels be-tween people who did and did not suffer from de-mentia were calculated using a standardized meandifference (SMD).13 Tests of heterogeneity were con-ducted with the Q statistic that is distributed as a χ2

variate under the assumption of homogeneity of ef-fect sizes. Between-study heterogeneity was assessedwith the I2 statistic. As a guide, I2 values of 25%may be considered low, 50% moderate, and 75% high.Funnel plots and statistical tests for funnel plot asym-metry were performed to test for evidence of publica-tion bias.

For models with considerable heterogeneity, meta-regression was performed to identify patient- andtrial related factors, which might contribute to theheterogeneity of effect sizes.14 Trial-related factorsincluded mean levels of creatinine, folate, VitaminB-12, and cobalamin, whereas patient-related factors

were mean age, mean years of education, proportionof men, and proportion of smokers. Mixed-effectsmeta-regression was used in the consideration ofthe presence of “residual heterogeneity.”14 Theregression coefficients and the associated z valuesand p values were reported in the meta-regressionanalysis. An R2 coefficient was also calculated toindicate the percentage of variance explained by themoderators. When the calculated R2 was negative, itwas truncated to zero.15

The OR and 95% CI for dementia in high-Hct levelgroups were obtained directly from the studies. Therandom-effects model was chosen to calculate an av-erage OR across relevant studies based on the testsfor heterogeneity.

RESULTS

From an initial 1,432 potentially relevant articles,63 articles met our inclusion criteria, of which, 48did not report relevant data to calculate effect size.Finally, we included 17 articles in our analysis (Fig. 1).

FIGURE 1. Flowchart Describing the Process of StudySelection




Of these, four were prospective studies that included2,631 participants in total, of which 155 of them devel-oped dementia. The remaining 13 articles were cross-sectional studies, comprising 1,256 dementia casesand 2,235 controls.

Majority of the studies were from western coun-tries and characteristics of the populations studiedare shown in Table 1. Generally, studies includedin the meta-analysis clearly reported informationabout recruitment, selection of subjects, and analyses.Fourteen16–29 of the 17 studies (4,009 participants)provided sufficient data to allow between-groupcomparisons of serum Hct levels in populations withdementia (AD and VaD) and without dementia. Fourstudies29–32 reported the ORs of developing demen-tia in high- versus low-Hct groups. Controls in case-control studies were chosen at random from thegeneral population. Response rates were not reportedfor most case-control studies, and four studies men-tioned blindness of Hct status to assessors of cog-nitive function16, 19,20,30 Seven studies16,17,19,21,26,29,30

reported the allele frequency of Apolipoprotein E e4+

among their participants. An average allele frequencyof 0.453 was found among 1,681 patients with AD,0.460 among 66 patients with VaD, and 0.193 among485 control participants.

Forrest Plots

Figure 2A shows the results of the 12 studies thatcompared the Hct levels between patients suffer-ing from AD and controls.16,18–26,28,29 The patientswith AD were significantly higher in their Hct levelthan control participants (pooled SMD with random-effects model: 0.59, 95% CI: 0.38–0.80, z = 5.43, p<0.001). A significant level of between-study het-erogeneity was found (τ 2 = 0.11, Q = 52.70, df =11, p <0.001, 12 = 79.1). When we undertook meta-regression to explore the impact of our a priorisources of heterogeneity, we found significant effectsof mean age (B = 0.07, z = 3.16, p = 0.002, R2 = 0.23).The other moderators (mean years of education, pro-portion of men, mean creatinine level, mean folatelevel, mean level of Vitamin B-12, and proportion ofsmokers) were not significant (R2 = 0) (Table 2).

Six studies were included that compared the Hctlevels between patients suffering from VaD andcontrols.18,20,22, 24–26 Results revealed that patientswith VaD were significantly higher in their Hct level

than control participants (pooled SMD with random-effects model: 1.30, 95% CI: 0.75–1.84, z = 4.65, p<0.001) (Fig. 2B). There was a significant level ofbetween-study heterogeneity (τ 2 = 0.38, Q = 32.11,df = 5, p <0.001, I2 = 84.4). We found significant ef-fects of mean age (B = 0.14, z = 3.66, p <0.001, R2 =0.15) and mean folate levels (B = −0.30, z = −4.32, p<0.001, R2 = 0.28) on heterogeneity. The other mod-erators (proportion of men and mean Vitamin B-12level) were either nonsignificant or did not contributeto the R2 (R2 = 0) (Table 3).

Figure 2C shows eight studies that comparedHct levels between patients suffering from AD andVaD.17,18,20,22, 24–27 It was found that patients withVaD had significantly higher Hct levels than patientswith AD (pooled SMD with random-effects model:0.48, 95% CI: 0.23–0.73, z = 3.72, p <0.001). A moder-ate level of between-study heterogeneity was found(τ 2 = 0.08, Q = 17.11, df = 7, p = 0.017, I2 = 59.1).We found significant effects of proportion of men(B = 0.04, z = 3.20, p = 0.001, R2 = 0.80) and meanfolate levels (B = −0.09, z = −2.45, p = 0.014, R2 =0.54) on heterogeneity. The other moderators (meanage, mean years of education, and mean Vitamin B-12level) were either nonsignificant or contributed noth-ing to the R2 (R2 = 0) (Table 4).

Four of the included studies examined the re-lationship between high-Hct status and dementia/cognitive decline as categorical variables and pro-vided sufficient data to calculate OR (Fig. 3). Theseprospective studies involved 2,631 participants.29–32

Follow-up was done over 24 months, and cognitivedeterioration was ascertained by a 3-point drop onthe Mini-Mental State Examination. There was no as-sociation between high Hct level and the risk of de-veloping dementia (random-effects pooled OR: 1.34,95% CI: 0.94–1.91, z = 1.63, p = 0.104). The degreeof between-study heterogeneity was not statisticallysignificant (τ 2 = 0.05, Q = 4.74, df = 3, p = 0.192,I2 = 36.7).

Publication Bias

We also tested for the presence of publication biasusing funnel plots and Egger’s regression test.33 Pub-lication bias was evident in the AD versus controlsanalysis (intercept = 4.32, 95% CI: −0.21 to 8.85, τ =2.16, df = 9, p = 0.030), and based on the classic fail-safe test, 334 missing studies are required for every



Ho et al.

TA

BLE

1.

Stu

dy

Des

ign

and

Bas

elin

eC

har

acte

rist

ics

of

Stu

die

sIn

clu

ded

Stu

dy

Stu

dy

Nam

eo

rLo

cati

on

Stu

dy

Des

ign

Po

pu

lati

on

Cas

eD

efin

itio

nC

on

tro

ls

Ob

serv

atio

nP

erio

d,

mo

nth

s(S

D)

Met

ho

ds

of

Co

gnit

ive

Ass

essm

ent

Mea

nA

ge,

yea

rs(S

D);

Mal

e(%

)M

ean

Hct

,μ

mo

l/L

(SD

)

Hct

Ter

tile

,μ

mo

l/L;

Od

ds

Rat

io(9

5%

CI)

Cla

rke

etal

.16O

xfo

rdsh

ire

Hea

lth

Au

tho

rity

,U

K

Cas

e–co

ntr

ol

Pat

ien

ts≥5

5ye

ars

fro

mth

eO

PT

IMA

stu

dy

CER

AD

and

his

tolo

gica

llyd

iagn

ose

dA

D(n

=76

)

Eld

erly

wit

ho

ut

mem

ory

imp

airm

ent

(n=

108)

NA

CER

AD

,CA

MC

OG

,M

MSE

76.7

(8.0

);37

.0%

AD

:16.

3(7

.4)

Hcy

11.1

–14.

0;O

R1.

1(0

.7–1

.9)

72.8

(8.8

);43

.0%

C:3

1.2

(4.0

)H

cy>

14.0

;OR

2.0

(1.1

–3.4

)Le

hm

ann

etal

.17Sw

eden

Cas

e–co

ntr

ol

Ou

tpat

ien

tsfr

om

am

emo

rycl

inic

DSM

-IVan

dN

INC

DS-

AD

RD

AA

D≥6

5ye

ars

(n=

54)

and

NIN

DS-

AIR

ENV

aD(n

=66

)

NA

NA

MM

SE,

NIN

CD

S-A

DR

DA

,N

IND

S-A

IREN

,n

euro

logi

can

dn

euro

psy

cho

-lo

gica

ltes

tin

g

75.8

(5.0

);38

.9%

AD

:14.

6(4

.6)

NA

76.7

(6.3

);54

.5%

VaD

:17.

1(5

.8)

Bo

ttig

lieri

etal

.18It

aly

Cas

e–co

ntr

ol

Co

nse

cuti

vep

atie

nts

fro

ma

mem

ory

clin

ic

Pre

vio

usl

yd

iagn

ose

dA

D(n

=48

)an

dV

aD(n

=7)

Eld

erly

pat

ien

tsfr

om

surg

ery,

MM

SEsc

ore

>26

(n=

22)

NA

MM

SE71

.0(8

.5);

33.3

%A

D:1

2.4

(11.

1)N

A72

.9(6

.2);

57.1

%V

aD:1

9.6

(16.

8)75

.6(6

.2);

31.8

%C

:11.

7(6

.2)

Ho

gerv

ors

tet

al.19

Ox

ford

shir

e,U

KC

ase–

con

tro

lP

arti

cip

ants

fro

mth

eO

PT

IMA

stu

dy

NIN

CD

S-A

DR

DA

AD

(n=

137)

Eld

erly

>50

year

sw

ith

ou

tco

gnit

ive

dys

fun

ctio

n,

MM

SE>

24(n

=27

7)

NA

MM

SE,C

AM

CO

G73

.9(9

.0);

40.0

%A

D:1

4.7

(4.9

)N

A

73.3

(7.7

);50

.0%

C:1

2.8

(3.9

)

Nag

gaet

al.20

Swed

enC

ase–

con

tro

lC

on

secu

tive

inp

atie

nts

and

ou

tpat

ien

tsfr

om

age

riat

ric

clin

ic

ICD

-10

AD

(n=

47)

and

VaD

(n=

59)

Ran

do

mse

lect

edel

der

ly≥6

0ye

ars

fro

ma

gast

riti

ssc

reen

ing

(n=

101)

NA

MM

SE74

.7(7

.3);

42.6

%A

D:1

6.5

(6.4

)N

A78

.4(5

.4);

37.3

%V

aD:1

7.3

(5.6

)69

.0(5

.8);

51.5

%C

:12.

9(4

.2)

Rel

iga

etal

.21P

ola

nd

Cas

e–co

ntr

ol

Pat

ien

tsfr

om

anA

lzh

eim

er’s

Day

Clin

ic

DSM

-IVan

dN

INC

DS-

AD

RD

Ad

emen

tia

(n=

99)

Eld

erly

wit

ho

ut

cogn

itiv

eim

pai

rmen

t(n

=10

0)

NA

NIN

CD

S-A

DR

DA

,D

SM-IV

74.2

(6.3

);A

D:1

8.0

(9.9

)N

A71

.2(6

.0);

C:1

4.4

(4.5

)

Ko

seo

glu

etal

.22T

urk

eyC

ase–

con

tro

lP

atie

nts

fro

ma

un

iver

sity

Neu

rolo

gyd

epar

tmen

tw

ith

AD

or

VaD

NIN

CD

S-A

DR

DA

AD

(n=

51);

NIN

DS-

AIR

ENV

aD(n

=67

)

Ran

do

mly

sele

cted

peo

ple

wit

hn

od

emen

tia

(n=

40)

NA

NIN

DS-

AIR

EN,

NIN

CD

S-A

DR

DA

,M

MSE

78.3

(4.1

);70

.0%

AD

:14.

2(3

.0)

NA

80.0

(4.8

);62

.7%

VaD

:18.

6(4

.2)

76.1

(3.9

);42

.5%

C:1

0.3

(1.3

)79

.050

.0%

AB

C36

:11.

9




TA

BLE

1.

(Co

nti

nu

ed)

Stu

dy

Stu

dy

Nam

eo

rLo

cati

on

Stu

dy

Des

ign

Po

pu

lati

on

Cas

eD

efin

itio

nC

on

tro

ls

Ob

serv

atio

nP

erio

d,

mo

nth

s(S

D)

Met

ho

ds

of

Co

gnit

ive

Ass

essm

ent

Mea

nA

ge,

yea

rs(S

D);

Mal

e(%

)M

ean

Hct

,μ

mo

l/L

(SD

)

Hct

Ter

tile

,μ

mo

l/L;

Od

ds

Rat

io(9

5%

CI)

Mill

eret

al.23

CA

Cro

ss–s

ecti

on

alO

utp

atie

nts

fro

man

Alz

hei

mer

’sD

isea

seC

ente

r

NIN

CD

S-A

DR

DA

AD

(n=

32)

Eld

erly

wit

ho

ut

AD

or

maj

or

neu

rod

egen

era-

tive

dis

ease

s(n

=22

)

NA

MM

SE78

.0(7

.0);

40.6

%A

D:1

0.6

(2.0

)H

cy>

12.0

;OR

2.2

(0.3

–16.

0)75

.0(7

.0);

36.4

%C

:9.3

(2.2

)

Sto

rey

etal

.24N

YC

ross

–sec

tio

nal

Pat

ien

tsfr

om

anel

der

lyn

urs

ing

ho

me

MM

SEsc

ore

<20

(AD

n=

50)

and

his

tory

of

coro

nar

yar

tery

dis

ease

,st

roke

,or

per

iph

eral

arte

rial

dis

ease

(VaD

n=

50)

Peo

ple

wit

ho

ut

dem

enti

ao

rva

scu

lar

dis

ease

fro

mth

esa

me

po

pu

lati

on

(n=

50)

NA

No

ne

79.0

(10.

0);3

4.0%

AD

:14.

4(2

.7)

NA

78.0

(10.

0);3

4.0%

VaD

:15.

3(3

.0)

79.0

(9.0

);34

.0%

C:1

0.6

(3.2

)

Qu

adri

etal

.25Sw

itze

rlan

dC

ross

–sec

tio

nal

Co

nse

cuti

vep

atie

nts

old

er60

year

sfr

om

am

emo

rycl

inic

CER

AD

AD

(n=

74)

and

VaD

(n=

18)

Eld

erly

wit

ho

ut

cogn

itiv

eim

pai

rmen

tfr

om

the

sam

ep

op

ula

tio

n(n

=55

)

NA

CER

AD

,CD

RS,

NIN

CD

S-A

DR

DA

,M

MSE

79.1

(7.7

);34

.3%

AD

:16.

8(7

.0)

NA

80.5

(5.7

);38

.9%

VaD

:18.

9(7

.9)

75.6

(8.5

);38

.2%

C:1

4.6

(6.1

)

Folin et

al.26

Ital

yC

ross

–sec

tio

nal

Pat

ien

tsfr

om

anA

lzh

eim

erEv

alu

atio

nU

nit

AD

and

VaD

dia

gno

sed

by

neu

rolo

gist

san

dp

sych

iatr

ists

(AD

n=

79;

VaD

n=

13)

Peo

ple

wit

ho

ut

anac

tive

med

ical

his

tory

fro

mth

esa

me

po

pu

lati

on

(n=

24)

NA

MM

SE,C

DR

S80

.3(7

.1);

AD

:21.

0(7

.8)

NA

82.5

(4.8

);V

aD:2

4.5

(8.1

)71

.2(9

.7);

C:1

5.8

(5.6

)

Qu

adri

etal

.27Sw

itze

rlan

dC

ross

–sec

tio

nal

Co

nse

cuti

vep

atie

nts

fro

ma

mem

ory

clin

ic

CER

AD

and

NIN

CD

S-A

DR

DA

AD

(n=

111)

and

VaD

(n=

25)

NA

NA

NIN

CD

S-A

DR

DA

,C

ERA

D78

.9(7

.5);

33.3

%A

D:1

6.9

(7.3

)N

A81

.0(5

.2);

44.0

%V

aD:2

0.0

(8.1

)

Haa

net

al.28

CA

Cro

ss–s

ecti

on

alP

arti

cip

ants

fro

mth

eSa

cram

ento

Are

aLa

tin

oSt

ud

yo

nA

gin

g

DSM

-III

dem

enti

a,fa

ilure

of≥1

mem

ory

sub

test

on

neu

rop

sych

olo

gi-

calt

est

(n=

93)

Peo

ple

wit

ho

ut

dem

enti

afr

om

the

sam

ep

op

ula

tio

n(n

=1,

436)

NA

MM

SE,D

SM-II

IN

AA

D:1

1.7

(4.1

)N

A

C:1

0.7

(6.7

)



Ho et al.

TA

BLE

1.

(Co

nti

nu

ed)

Stu

dy

Stu

dy

Nam

eo

rLo

cati

on

Stu

dy

Des

ign

Po

pu

lati

on

Cas

eD

efin

itio

nC

on

tro

ls

Ob

serv

atio

nP

erio

d,

mo

nth

s(S

D)

Met

ho

ds

of

Co

gnit

ive

Ass

essm

ent

Mea

nA

ge,

yea

rs(S

D);

Mal

e(%

)M

ean

Hct

,μ

mo

l/L

(SD

)

Hct

Ter

tile

,μ

mo

l/L;

Od

ds

Rat

io(9

5%

CI)

Kal

mijn

etal

.30N

eth

erla

nd

sP

rosp

ecti

veP

arti

cip

ants

≥55

year

sfr

om

the

Ro

tter

dam

Stu

dy

To

tal(

n=

702)

;in

clu

din

gco

gnit

ive

dec

line

(n=

110)

NA

31M

MSE

69.9

(9.4

);36

.0%

14.7

(0.9

5)H

cy12

.9–1

5.7;

OR

1.1

(0.7

–1.9

)H

cy>

15.8

;OR

0.9

(0.5

–1.6

)D

ufo

uil

etal

.31Fr

ance

Pro

spec

tive

Par

tici

pan

tsfr

om

ap

op

ula

tio

nco

ho

rtfr

om

the

Epid

emio

logy

of

Vas

cula

rA

gein

gst

ud

y

N=

1,24

1N

A24

MM

SE(c

ogn

itiv

ed

eter

iora

tio

nw

ith

dec

reas

eo

f≥

3p

oin

ts),

WA

IS-R

67.0

(3.0

);41

.4%

12.2

(2.6

)H

cy10

.0–1

1.9;

OR

1.6

(0.7

–3.4

)H

cy12

.0–1

4.9;

OR

1.3

(0.5

–2.9

)H

cy>

15;O

R2.

8(1

.2–6

.2)

Hcy

>15

.8;O

R0.

9(0

.5–1

.6)

Ro

wan

etal

.32T

he

Un

ited

Kin

gdo

mP

rosp

ecti

ve3

mo

nth

sp

ost

stro

ke,

wel

l-rec

ove

red

,n

on

dem

ente

del

der

lyp

atie

nts

N=

170

NA

24C

AM

CO

G;D

SM-IV

No

ne

16.0

Hcy

>15

.0;O

R1.

1(0

.8–1

.7)

Kim et

al.29

Kw

angj

u,

Sou

thK

ore

a

Pro

spec

tive

Eld

erly

>65

year

sfr

om

ap

op

ula

tio

nsu

rvey

DSM

-IVd

emen

tia

(n=

45)

Co

mm

un

ity

resi

den

ts(n

=47

3)

28M

MSE

,CD

RS,

WH

OD

AS

II,G

MS

dep

ress

ion

74.3

(5.7

);36

.2%

AD

:14.

3(8

.1)

Hcy

>15

.0;O

R1.

6(0

.7–3

.5)

71.6

(4.9

);45

.1%

C:1

2.3

(5.2

)

Not

es:C

:con

trol

part

icip

ants

;CA

MC

OG

:Cam

brid

geC

ogni

tive

Exa

min

atio

n;C

DR

S:C

linic

alD

emen

tia

rati

ngsc

ale;

CE

RA

D:C

onso

rtiu

mto

Est

ablis

ha

Reg

istr

yfo

rA

lzhe

imer

’sD

isea

se;

DSM

-IV

:Dia

gnos

tic

and

Stat

isti

cal

Man

ual

ofM

enta

lD

isor

ders

-IV

;GM

S:G

eria

tric

Men

tal

Stat

esc

hed

ule;

Hct

:hom

ocys

tein

e;M

MSE

:Min

i-M

enta

lSt

ate

Exa

min

atio

n;N

INC

DS-

AD

RD

A:N

a-ti

onal

Inst

itut

eof

Neu

rolo

gica

land

Com

mun

icat

ive

Dis

ord

ers

and

Stro

kean

dth

eA

lzhe

imer

’sD

isea

sean

dR

elat

edD

isor

der

sA

ssoc

iati

on;N

IND

S-A

IRE

N:N

atio

nalI

nsti

tute

ofN

euro

logi

cal

Dis

ord

ers

and

Stro

ke-A

ssoc

iati

onIn

tern

atio

nale

pour

laR

eche

rche

etl’E

nsei

gnem

ente

nN

euro

scie

nces

;OPT

IMA

:Oxf

ord

Proj

ectt

oIn

vest

igat

eM

emor

yan

dA

gein

g;W

AIS

-R:W

echs

ler

Ad

ult

Inte

llige

nce

Scal

e-R

evis

ed;W

HO

DA

S-II

:Wor

ldH

ealt

hO

rgan

izat

ion

Dis

abili

tyA

sses

smen

tSch

edul

eII

.




FIGURE 2. Meta-Analysis of Case-Control Studies onAssociation of Hct Levels and Dementia

Standardized mean differences of mean plasma Hct levels of [A] pa-tients with AD compared with healthy control; [B] patients with VaDcompared with healthy controls; and [C] patients with VaD comparedwith patients with AD.

observed study to nullify the effect. Publication biaswas not evident in other comparisons.

DISCUSSION

This study represents, to our knowledge, the firstmeta-analysis of Hct and dementia based on a broad

range of studies involving 6,122 participants. Thismeta-analysis shows that high-Hct levels were foundin individuals with AD and VaD, when comparedwith controls, but high baseline Hct level was notassociated with subsequent risk of developing de-mentia. Our findings do not support a causal rela-tionship between high-Hct level and the risk of de-veloping dementia. We also found a stronger positiveassociation between hyperhomocysteinemia andVAD, when compared with AD, which is consistentwith the finding that patients with vascular diseaseshave significantly higher levels of Hct.34

Strengths

The strengths of this review include an a priorisearch strategy, duplicate study selection and data ex-traction, a comprehensive literature review, and thereporting of inclusion and exclusion criteria.

In addition, one of the key elements to a well-designed case-control study is to have good responserate. The mean response rate reported in the case-control studies31,32 was 83.44%. Thus, it is unlikelythat nonresponse could bias the results appreciably.For cohort studies, completeness of follow-up is ofgreater importance than the response rate at entryinto the study. The loss to follow-up reported in thestudies was low (<15%).

Limitations

We cannot ignore the limitations of performing ameta-analysis on observational studies. The largestbody of evidence for this review comes from epide-miologically weaker cross-sectional and case–controlstudies, and the association between high-Hct levelsand dementia may be due to reverse causality. Otherconfounding factors such as health and life style fac-tors associated with Hct may explain the associationsbetween high-Hct levels and dementia or cognitivedecline in cross-sectional studies. For example, high-Hct levels may result from poor nutrition or poorgeneral health and be associated with less physicalactivity or frailty. The lack of reported informationin the included studies on participants taking Vita-min B supplements may confound the results. Asthe majority of studies were conducted in westerncountries, cultural and ethnic differences in life styleand genetic factors may also have an impact on the



Ho et al.

TABLE 2. Results of Random-Effects Meta-Regression for Patients With AD Versus Healthy Controls

No. of UnivariatePredictors Studies Used Coefficient Z p t2 Estimate R2

Age, years 11 0.068 3.160 0.002 0.072 0.233Education, years 4 0.029 1.123 0.261 0.357 0.000Gender, males 9 −0.005 −0.397 0.692 0.152 0.000Creatinine, μmol/L 4 0.001 0.126 0.900 0.000 0.000Folate, nmol/L 5 −0.023 −1.501 0.133 0.248 0.000Vitamin B-12, pmol/L 7 −0.001 −0.650 0.516 0.159 0.000Smoker 3 −0.001 −0.175 0.861 0.001 0.000

Notes: Range of moderators: age, years (70.81–78.18); education, years (2.74–10.95); male gender (0.33–0.49); creatinine, μmol/L (70.72–95.12); folate, nmol/L (9.04–24.86); Vitamin B-12, pmol/L (237.30–380.73); and smoker (0.12–0.42).

TABLE 3. Results of Random-Effects Meta-Regression for Patients With VaD Versus Healthy Controls

No. of UnivariatePredictors studies used Coefficient Z p t2 Estimate R2

Age, years 6 0.144 3.659 <0.001 0.321 0.151Gender, males 5 0.040 2.685 0.007 0.473 0.000Folate, nmol/L 3 −0.297 −4.319 <0.001 0.840 0.275Vitamin B-12, pmol/L 3 −0.002 −0.463 0.643 1.782 0.000

Notes: Range of moderators: age, years (72.47–78.54); male gender (0.34–0.55); folate, nmol/L (10.15–15.54); and Vitamin B-12, pmol/L(289.81–377.62).

TABLE 4. Results of Random-Effects Meta-Regression for Patients With VaD Versus Patients With AD

No. of UnivariatePredictors Studies Used Coefficient Z p t2 Estimate R2

Age, years 8 0.035 0.843 0.399 0.087 0.000Education, years 3 −1.161 −1.938 0.053 0.311 0.000Gender, males 7 0.037 3.202 0.001 0.019 0.794Folate, nmol/L 5 −0.090 −2.451 0.014 0.047 0.538Vitamin B-12, pmol/L 5 −0.002 −1.332 0.183 0.143 0.000

Notes: Range of moderators: age, years (71.24–80.61); education, years (7.97–8.45); male gender (0.34–0.53); folate, nmol/L (8.34–15.96);and Vitamin B-12, pmol/L (251.97–445.66).

observed associations. Another limitation of thisstudy is that we were only able to include a smallnumber of moderators in the meta-regression anal-yses because of the limited number of studies thatprovided data of trial- and patient-related modera-tors. In addition, we did not find an overall associ-ation linking high-Hct levels and the risk of devel-oping dementia in prospective studies. This could beattributed to a lack of statistical power, given thatonly four prospective studies were included in theOR analysis. Finally, we were only able to test thelinear effect of the mean age on the associations be-tween Hct level and cognitive decline. It has beenobserved that the relationship between the demen-tia prevalence and age may be exponential ratherthan linear.35 Because the mean age in the data was

restricted from 70 to 80 years, this study could notverify whether the effect of age on the associationsbetween Hct level and cognitive decline is exponen-tial. Further studies may address how age affects theassociations between Hct level and cognitive decline.

Clinical Implications

Complementary Evidence From Gene-Association Stud-ies. The possibility that high-Hct levels may play ina role in the pathogenesis of dementia cannot becompletely discarded. Preliminary evidence of a di-rection of causality comes from the association be-tween impairments in Hct metabolism and demen-tia.36 Evidence of hyperhomocysteinemia is relatedto the progression of dementia, and increasing sever-ity in dementia is strengthened by the relationship




FIGURE 3. Odds Ratio of Developing Dementia in Individu-als With High-Hct Levels, When Compared WithIndividuals With Low-Hct Levels

between Hct and the homozygous C677T mutationof the methylene tetrahydrofolate reductase gene.18

Evidence of Therapeutic Benefit From Supplements thatLower Hct Level. Research in Hct-lowering studieswas mainly conducted in patients with myocardialinfarction and stroke. The effectiveness of using vi-tamin therapy to lower Hct in preventing vasculardisease is controversial.37–39 Spence40 suggested thatHct-lowering therapy could be more effective for theprevention of cerebral infarction by decreasing Hctvia reducing thrombosis and embolic events. Sim-ilarly, research studying the effect of Hct-loweringtherapy on prevention of dementia also yields con-flicting results. Aisen et al.41 conducted a random-ized controlled trial comparing the high dose of Vi-tamin B supplementation versus placebo treatmentin 340 patients with AD in the United States andfound no beneficial effect on the rate of cognitive

decline after 18 months despite significant reductionin Hct levels. This study also yielded an unexpectedfinding where higher quantity of adverse events in-volving depression was found in the high-dose Bvitamin supplementation group. This negative find-ing could be related to the fact that these patients hadbeen consuming grain products with folic acid sup-plementation before commencement of the trial. Fur-thermore, it may be too late for Hct-lowering ther-apy to yield significant results in patients who havealready developed AD. The effect of Hct-loweringtherapy requires further evaluation in populationswithout folate supplementation in their food, in non-dementia populations, and in patients with othertypes of dementia.

Future Research

Although this meta-analysis found that patientswith dementia have higher Hct levels compared withcontrols, results failed to show the association be-tween high-Hct levels and the subsequent risk of de-veloping dementia, the role of Hct in cognitive de-cline, and dementia can only be adequately answeredfrom well-designed cohort studies that have more de-tailed neuropsychological measurements such as forexecutive functioning. It is critical to study the rela-tionship between Hct and cognitive decline startingfrom midlife, and this review highlights the need formore data from future longitudinal studies involv-ing several occasions of measurements of Hct leveland cognitive impairment from midlife to late adult-hood. A more in-depth understanding of the under-lying mechanisms linking Hct to cognitive decline isrequired.

References

1. Ferri CP, Prince M, Brayne C, et al; Alzheimer’s DiseaseInternational: Global prevalence of dementia: a Delphi consen-sus study. Lancet 2005; 366:2112–2117

2. McCaddon A: Homocysteine and cognition—a historical perspec-tive. J Alzheimers Dis 2006; 9:361–380

3. Garcia A, Zanibbi K: Homocysteine and cognitive function inelderly people. Can Med Assoc J 2004; 171:897–904

4. Barberger-Gateau P, Raffaitin C, Letenneur L, et al: Dietarypatterns and risk of dementia: the Three-City Cohort Study.Neurology 2007; 69:1921–1930

5. Del Parigi A, Panza F, Capurso C, et al: Nutritional factors, cogni-tive decline, and dementia. Brain Res Bull 2006; 69:1–19

6. Farias ST, Mungas D, Reed BR, et al: Progression of mild cogni-tive impairment to dementia in clinic- versus community-basedcohorts. Arch Neurol 2009; 66:1151–1157

7. Berlin JA: Invited commentary. Am J Epidemiol 1995; 142:3833878. Anstey KJ, Lipnicki DM, Low LF: Cholesterol as a risk factor for

dementia and cognitive decline: a systematic review of prospec-tive studies with meta-analysis. Am J Geriatr Psychiatry 2008; 16:343–354

9. Stroup DF, Berlin JA, Morton SC, et al: Meta-analysis of ob-servational studies in epidemiology: a proposal for reporting.Meta- Analysis of Observational Studies in Epidemiology (MOOSE)Group. J Am Med Assoc 2000; 283:2008–2012



Ho et al.

10. NHS Centre for Reviews and Dissemination. Undertaking System-atic Reviews of Research on Effectiveness: CRD Report 4, 2nd ed.York, University of York, 2001

11. R: A Language and Environment for Statistical Computing[Computer Program]. Vienna, R Development Core Team, 2010

12. Metafor: Meta-Analysis Package for R [Computer Program].Version 0.5–7. Netherlands, Viechtbauer W, 2009

13. Sutton AJ, Abrams KR, Jones DR, et al: Systematic reviews of trialsand other studies. Health Technol Assess 1999; 2:1–276

14. Thompson SG, Higgins JP: How should meta-regression analysesbe undertaken and interpreted? Stat Med 2002; 21:1559–1573

15. Raudenbush SW: Analyzing effect sizes: random-effects models,in The Handbook of Research Synthesis and Meta-Analysis, 2nded. Edited by Cooper H, Hedges LV, Valentine JC. New York,Russell Sage Foundation, 2009, pp 295–315

16. Clarke R, Smith AD, Jobst KA: Folate, vitamin B12, and serumtotal homocysteine levels in confirmed Alzheimer disease. ArchNeurol 1998; 55:1449–1455

17. Lehmann M, Gottfries CG, Regland B: Identification of cogni-tive impairment in the elderly: homocysteine is an early marker.Dement Geriatr Cogn Disord 1999; 10:12–20

18. Bottiglieri T, Parnetti L, Arning E, et al: Plasma total homocysteinelevels and the C677T mutation in the methylenetetrahydrofolatereductase (MTHFR) gene: a study in an Italian population withdementia. Mech Ageing Dev 2001; 122:2013–2023

19. Hogervorst E, Ribeiro HM, Molyneux A, et al: Plasma homocys-teine levels, cerebrovascular risk factors, and cerebral white mat-ter changes (Leukoaraiosis) in patients with Alzheimer disease.Arch Neurol 2002; 59:787–793

20. Nagga K, Rajani R, Mardh E, et al: Cobalamin, folate, methyma-lonic acid, homocysteine, and gastritis markers in dementia.Dement Geriatr Cogn Disord 2003; 16:269–275

21. Religa D, Styczynska M, Peplonska B, et al: Homocysteine,apolopoproteine E and methylenetetrahydrofolate reductase inAlzheimer’s disease and mild cognitive impairment. DementGeriatr Cogn Disord 2003; 16:64 -70

22. Koaseoglu E, Karaman Y: Relations between homocysteine, folateand vitamin B12 in vascular dementia and in Alzheimer disease.Clin Biochem 2007; 40:859–863

23. Miller JW, Green R, Mungas DM, et al: Homocysteine,vitamin B6, and vascular disease in AD patients. Neurology 2002;58:1471–1475

24. Storey SG, Suryadevara V, Aronow WS: Association of plasmahomocysteine in elderly persons with atherosclerotic vasculardisease and dementia, atherosclerotic vascular disease withoutdementia, dementia without atherosclerotic vascular disease, andno dementia or atherosclerotic vascular disease. J Gerontol 2003;58:1135–1136

25. Quadri P, Fragiacomo C, Pezzati R, et al: Homocysteine, folate,and vitamin B-12 in mild cognitive impairment, Alzheimer disease,and vascular dementia. Am J Clin Nutr 2004; 80:114–122

26. Folin M, Baiguera S, Gallucci M, et al: A cross-sectional study ofhomocysteine, NO-levels, and CT-findings in Alzheimer dementia,vascular dementia and controls. Biogerontology 2005; 6:255–260

27. Quadri P, Fragiacomo C, Pezzati R, et al: Homocysteine and Bvitamins in mild cognitive impairment and dementia. Clin ChemLab Med 2005; 43:1096–1100

28. Haan MN, Miller JW, Aiello AE, et al: Homocysteine, B vitamins,and the incidence of dementia and cognitive impairment: resultsfrom the Sacramento Area Latino Study on Aging. Am J Clin Nutr2007; 85:511–517

29. Kim JM, Stewart R, Kim SW, et al: Changes in folate, vitamin B12and homocysteine associated with incident dementia. J NeurolNeurosurg Psychiatry 2008; 79:864–868

30. Kalmijn S, Launer LJ, Lindemans J, et al: Total homocysteine andcognitive decline in a community-based sample of elderly sub-jects. Am J Epidemiol 1999; 150:283–289

31. Dufouil C, Alperovitch A, Ducros V: Homocysteine, white matterhyperintensities, and cognition in healthy elderly people. AnnNeurol 2003; 53:214–221

32. Rowan EN, Dickinson HO, Stephens S, et al: Homocysteine andpost-stroke cognitive decline. Age Ageing 2007; 36:339–343

33. Egger M, Davey Smith G, Schneider M, et al: Bias in meta-analysisdetected by a simple graphical test. BMJ 1997; 315:629–634

34. Khandanpour N, Loke YK, Meyer FJ, et al: Homocysteine andperipheral arterial disease: systematic review and meta-analysis.Eur J Vasc Endovasc Surg 2009; 38:316–322

35. Ritchie K, Kildea D: Is senile dementia “age-related” or “ageing-related”? Evidence from meta-analysis of dementia prevalence inthe oldest old. Lancet 1995; 346:931–934

36. Postiglione A, Milan G, Ruocco A, et al: Plasma folate, vitaminB12, and total homocysteine and homozygosity for the C677Tmutation of the 5,10-methylene tetrahydrofolate reductase genein patients with Alzheimer’s dementia: a case-control study.Gerontology 2001; 47:324–329

37. Toole JF, Malinow MR, Chambless LE, et al: Lowering plasma to-tal homocysteine to prevent recurrent stroke, myocardial infarc-tion, and death in ischemic stroke patients: results of the VitaminIntervention for Stroke Prevention (VISP) Randomized Trial. J AmMed Assoc 2004; 291:565–575

38. Bonaa KH, Njolstad I, Ueland PM, et al: Homocysteine loweringand cardiovascular events after acute myocardial infarction. NEngl J Med 2006; 354:1578–1588

39. Lonn E, Yusuf S, Arnold MJ, et al: Homocysteine lowering withfolic acid and B vitamins in vascular disease. N Engl J Med 2006;354:1567–1577

40. Spence JD: Homocysteine-lowering therapy: a role in stroke pre-vention? Lancet Neurol 2007; 6:830–838

41. Aisen PS, Schneider LS, Sano M et al: High-dose B vitamin supple-mentation and cognitive decline in Alzheimer disease: a random-ized controlled trial. Alzheimer Disease Cooperative Study. JAMA2008; 300:1774–1783



is high homocysteine level a risk factor for cognitive decline in elderly? a systematic review,...

Documents