Does tailoring make a difference? A systematic review of the long-term effectiveness of tailored...

17
Does tailoring make a difference? A systematic review of the long-term effectiveness of tailored nutrition education for adults Helen C Eyles and Cliona Ni Mhurchu Tailoring individualizes information to the receiver and provides a potential strategy for improving dietary intakes. The present systematic review summarizes evidence for the long-term (6 months) effectiveness of tailored nutrition education for adults and includes priority population groups. Key electronic databases and relevant bibliographies were searched for trials measuring the following outcomes: nutrition-related health behaviors (e.g., dietary intake and food purchases) and anthropometric measures. Data synthesis was comprised of meta-analysis (for 15 trials including all population groups) and narrative review (for five trials of priority population groups). Overall, the quality of the studies was moderate to good. Tailored nutrition education was found to be a promising strategy for improving the diets of adults (including those in priority population groups) over the long term. However, future studies should ensure adequate reporting of research design and methods and reduce the chances of false-positive findings by using more objective measures of diet, clearly identifying the primary study outcome, and concentrating on outcomes most relevant to nutrition-related disease.© 2009 International Life Sciences Institute INTRODUCTION The impact of nutrition on health is substantial, with high blood pressure, high cholesterol, high body mass index, and low fruit and vegetable intake accounting for 30% of disability adjusted life years (DALYs) worldwide each year. 1 Risk of nutrition-related disease is also influenced by socioeconomic and cultural factors. Those on low incomes 2–5 or belonging to priority ethnic groups 6-8 are less likely to consume healthy diets and are thus at greater risk of nutrition-related disease. Nutrition knowledge is important for improving nutrition-related behaviors 9,10 ; however, the effectiveness of nutrition education interventions that aim to improve knowledge may be dependent upon relevance to the indi- vidual. Many people perceive their diet to be healthier than it is in reality, 11,12 and they may disregard generic nutrition education messages because they fail to find relevance in them. One potential way to overcome this is to provide tailored (personalized) feedback on diet. 13 In a summary of reviews on effecting dietary change, 14 tailoring was found to be a key component of effective dietary interventions. However, only one sys- tematic review to date has focused specifically on the efficacy of tailored nutrition education (although this review also included physical activity). 15 In that review, Kroeze et al. 15 included randomized controlled trials of computer-tailored interventions that delivered education in a “non-personal” way. The trials examined nutrition- related behaviors in adults and were published between 1965 and September 2004. 15 Six trials reported effects over the long term (6 months). A narrative summary suggested tailoring had no significant impact on total fat (n = 2 studies) or calcium intake (n = 1 study). However, tailored nutrition education was found to be significantly more effective than generic and no nutrition Affiliation: HC Eyles and CN Mhurchu are with the Clinical Trials Research Unit, School of Population Health, University of Auckland, New Zealand. Corresponding author: Helen Eyles, Clinical Trials Research Unit, The University of Auckland, Private Bag 92019, Auckland, New Zealand, Phone: +64 9 373 7599 ext. 84658, Fax: +64 9 373 1710, E-mail: [email protected] Key words: tailoring, nutrition therapy, long-term effect, adult, health priorities Nutrition SciencePolicy doi:10.1111/j.1753-4887.2009.00219.x Nutrition Reviews® Vol. 67(8):464–480 464

Transcript of Does tailoring make a difference? A systematic review of the long-term effectiveness of tailored...

Does tailoring make a difference? A systematic review of thelong-term effectiveness of tailored nutrition educationfor adults

Helen C Eyles and Cliona Ni Mhurchu

Tailoring individualizes information to the receiver and provides a potential strategyfor improving dietary intakes. The present systematic review summarizes evidencefor the long-term (�6 months) effectiveness of tailored nutrition education foradults and includes priority population groups. Key electronic databases andrelevant bibliographies were searched for trials measuring the following outcomes:nutrition-related health behaviors (e.g., dietary intake and food purchases) andanthropometric measures. Data synthesis was comprised of meta-analysis (for 15trials including all population groups) and narrative review (for five trials of prioritypopulation groups). Overall, the quality of the studies was moderate to good.Tailored nutrition education was found to be a promising strategy for improving thediets of adults (including those in priority population groups) over the long term.However, future studies should ensure adequate reporting of research design andmethods and reduce the chances of false-positive findings by using more objectivemeasures of diet, clearly identifying the primary study outcome, and concentratingon outcomes most relevant to nutrition-related disease.nure_219 464..480

© 2009 International Life Sciences Institute

INTRODUCTION

The impact of nutrition on health is substantial, with highblood pressure, high cholesterol, high body mass index,and low fruit and vegetable intake accounting for 30% ofdisability adjusted life years (DALYs) worldwide eachyear.1 Risk of nutrition-related disease is also influencedby socioeconomic and cultural factors. Those on lowincomes2–5 or belonging to priority ethnic groups6-8 areless likely to consume healthy diets and are thus at greaterrisk of nutrition-related disease.

Nutrition knowledge is important for improvingnutrition-related behaviors9,10; however, the effectivenessof nutrition education interventions that aim to improveknowledge may be dependent upon relevance to the indi-vidual. Many people perceive their diet to be healthierthan it is in reality,11,12 and they may disregard genericnutrition education messages because they fail to find

relevance in them. One potential way to overcome this isto provide tailored (personalized) feedback on diet.13

In a summary of reviews on effecting dietarychange,14 tailoring was found to be a key component ofeffective dietary interventions. However, only one sys-tematic review to date has focused specifically on theefficacy of tailored nutrition education (although thisreview also included physical activity).15 In that review,Kroeze et al.15 included randomized controlled trials ofcomputer-tailored interventions that delivered educationin a “non-personal” way. The trials examined nutrition-related behaviors in adults and were published between1965 and September 2004.15 Six trials reported effectsover the long term (�6 months). A narrative summarysuggested tailoring had no significant impact ontotal fat (n = 2 studies) or calcium intake (n = 1 study).However, tailored nutrition education was found to besignificantly more effective than generic and no nutrition

Affiliation: HC Eyles and CN Mhurchu are with the Clinical Trials Research Unit, School of Population Health, University of Auckland, NewZealand.

Corresponding author: Helen Eyles, Clinical Trials Research Unit, The University of Auckland, Private Bag 92019, Auckland, New Zealand,Phone: +64 9 373 7599 ext. 84658, Fax: +64 9 373 1710, E-mail: [email protected]

Key words: tailoring, nutrition therapy, long-term effect, adult, health priorities

Nutrition Science↔Policy

doi:10.1111/j.1753-4887.2009.00219.xNutrition Reviews® Vol. 67(8):464–480464

education (combined) with regards to daily intake offruits and vegetables (n = 2 studies) and weight loss (n = 1study).15 The review did not explore evidence for theeffectiveness of tailored nutrition education for priorityethnic and low-income groups.15 However, in order totackle inequalities in nutrition-related health, the effec-tiveness of strategies designed to change behavior in thesegroups must be considered.

The aim of the current review was to update andevaluate the long-term effectiveness of tailored nutritioneducation for adults. Specific objectives were: 1) to deter-mine whether tailored nutrition education is effective forimproving diet-related behaviors compared with genericnutrition education and compared with no nutrition edu-cation (control); and 2) to determine whether tailorednutrition education is effective for improving the diet-related behaviors of priority ethnic and low-incomegroups.

LITERATURE REVIEW METHODS

This review was conducted using a protocol broadlybased on the Cochrane guidelines.16

Selection of studies

Manuscripts eligible for inclusion included randomizedor quasi-randomized controlled trials employing a paral-lel design in which at least one group of participantsreceived tailored nutrition education and one groupreceived either generic and/or no nutrition education.The following definitions of tailored and generic educa-tion from Kreuter et al.13 were used: 1) tailored nutritioneducation: “. . . any combination of information orchange strategies intended to reach one specific person,based on characteristics that are unique to that person, arerelated to the outcome of interest, and have been derivedfrom individual assessment”13; 2) generic nutrition edu-cation: “[is intended to] reach some specific subgroup ofthe general population usually based on one or moredemographic characteristics shared by its members.”13

The tailored nutrition education intervention hadto include at least one print, e-mail, or other non-face-to-face delivery format. Outcomes included were: 1)nutrition-related health behaviors, which included foodand nutrient intake and purchases, and 2) anthropomet-ric outcomes such as change in body weight, BMI, orwaist circumference. Eligible trials had at least 6 monthsof follow-up and included male and/or female adults(�18 years of age) of all ethnicities and any health status.Studies published prior to 1990 were excluded (to focuson methods of tailoring relevant to current technology),as were non-English publications, or those that included a

tailored nutrition education intervention delivered onlyby face-to-face methods. Trials of priority ethnic groupswere those in which the majority (>50%) of participantswere of priority ethnicity (definitions determined byauthors), or in which a priority ethnic group was com-pared with another ethnic group. Similarly, trials of thoseon a low-income were those in which the majority(>50%) of participants were described as being on a lowincome (definitions determined by authors), or in whichlow-income participants were compared with those onhigher incomes.

Search strategy and data sources

Studies published between January 1990 and December2007 were identified through structured searches of elec-tronic databases Medline, Medline In-Process, PsychInfo,Cinahl, Eric, Embase, DARE, CDSR, Digital Abstracts,Science Citation Index, and PubMed. The followingsearch strategy was run across Medline and modifiedwhere necessary for other databases: (expand (exp) nutri-tion therapy OR exp diet OR exp eating OR exp feedingbehaviour OR exp food OR $nutrition.mp) AND (healtheducation OR education OR exp education, nonprofes-sional or education, public health professional OR publichealth OR exp health promotion OR exp primary healthcare OR exp dietary services OR personal health servicesOR preventive health services OR educat$.mp) AND (per-sonal$.mp OR tailor$.mp OR individual$.mp OR own.mpOR privat$.mp OR personal health services) AND (tolimit to trials including priority ethnic groups) (ethnicgroups OR oceanic ancestry group OR culture$.mp ORminorit$.mp OR maori$.mp OR pacific$.mp or eth-nic$.mp) AND (to limit to trials including low-incomegroups)(exp socioeconomic factors OR vulnerable popula-tions OR low pay$.mp OR low salar$.mp OR low-income$.mp OR disadvantage$.mp).

All references were entered into an Endnote libraryand titles, descriptors, and abstracts were reviewed. Full-text copies of all relevant or potentially relevant trialswere obtained for review. In addition, bibliographies ofall relevant publications were manually searched. Oneauthor (HE) independently reviewed the articles forinclusion and the second author (CNM) was available toresolve any doubts as to whether specific studies wereeligible.

Data extraction and synthesis

The following data were extracted for each study includedin the review: objectives, setting, participants, inclusionand exclusion criteria, intervention and control, durationof follow-up, and results. Additional data were soughtfrom the authors where necessary, and methodology

Nutrition Reviews® Vol. 67(8):464–480 465

papers were obtained where possible to complete missinginformation. Study quality was assessed using the follow-ing criteria: presence or absence of random allocation;concealed allocation; comparable groups at baseline(indicators of selection bias); percentage loss to follow-up(indicator of attrition bias); intention-to-treat analyses(to account for attrition bias); blinding (indicator of per-formance and detection bias), and a priori sample sizecalculation (indicator of precision). Using the qualitycriteria and a taxonomy adapted from the CochraneHandbook for Systematic Reviews of Interventions,17

a summary “risk of bias” score (high, medium, low) wasassigned to each study (Table 1).

Results from studies that included all populationgroups and that were sufficiently alike in terms of com-parison groups (i.e., either generic or no nutrition edu-cation) and outcomes were combined in meta-analyses.Data synthesis was carried out using Cochrane ReviewManager (RevMan) 5, with weighted mean differencesand 95% confidence intervals reported for continuousoutcomes. Standard deviations were calculated fromstandard errors where necessary. Statistical heterogeneitywas tested using the chi-square method. Heterogeneitywas assumed with P-values <10%.16 Where statistical het-erogeneity was present, a random effects (rather than afixed effects) model was used. For meta-analyses (wherepossible), a sensitivity analysis was completed to assessthe effects of trials carrying a high risk of bias. A funnelplot was created to assess for likelihood of publicationbias. Results from all population group studies that werenot sufficiently alike to be combined in meta-analyseswere excluded.

Results from studies of priority ethnic and low-income groups were combined in a narrative review (thesmall number of studies and heterogeneous outcomes inthese subgroups prohibited meta-analyses).

INCLUDED STUDIES

Identification and selection of studies is summarized inFigure 1. A total of 127 potentially eligible studies wereidentified. Following review of abstracts, 35 full-text

manuscripts were obtained, and 25 trials were ultimatelyfound to meet all eligibility criteria.18–42 The 25 trials wereobtained from the electronic databases Medline (n = 14)and Embase (n = 6), and from bibliographies of relevantpublications (n = 5).

The 25 eligible studies (including priority ethnicand low-income groups) were diverse in terms oftheir methodology and outcomes. Due to this heteroge-neity, it was possible to perform meta-analyses for onlytwo outcomes: daily servings of fruits and vegetablesand percentage of energy consumed from total fat.Fifteen18,19,21,22,25,27,31–33,35–40 of the 25 studies assessed oneor both outcomes, and were therefore included in thisreview. Ten studies did not assess daily servings of fruitsand vegetables or percentage of energy from total fat, andwere thus excluded. A further 10 studies assessed out-comes in addition to daily servings of fruits and veg-etables or percentage of energy from total fat. Theseadditional outcomes were also excluded. Study character-istics and results for the 20 studies with excludedoutcomes18–26,28,31,33,34,36–42 are summarized in Table 2.

Four studies focused on priority ethnic groups21,22,26,29

and one on a low-income group.33 These studies wereincluded; however, four of these five additional trialsoverlapped with those in the meta-analyses. Therefore,the total number of trials included in this review paperwas 16.

No studies compared priority ethnic with other ethnicgroups. Similarly, no studies compared those on a lowincome with those on a higher income. Figure 1 details theresults of the search strategy and the included studies. Acomplete table of study characteristics and results for all 25trials is available in Table S1,provided as supporting infor-mation in the online version of this article.

CHARACTERISTICS OF INCLUDED STUDIES

The characteristics of the 15 studies (n = 20,809 partici-pants) included in meta-analyses, the four trials (n = 5,981participants) included in the priority ethnic subgroup,andthe single trial (n = 2,042 participants) included in thelow-income subgroup are described below:

Table 1 Guide to assigning summary risk of bias scores.Risk of bias Interpretation Relationship to individual bias criteriaLow Possible bias, unlikely to seriously affect the

study resultsAll criteria met; if criteria not reported, study does

not drop to medium category unless random/concealed allocation criteria not reported

Medium Possible bias that raises some doubt about theresults

One or more criteria partially met

High Possible bias that seriously weakens confidencein the results

One or more criteria not met

Adapted from The Cochrane Handbook (2006)17.

Nutrition Reviews® Vol. 67(8):464–480466

All population groups (meta-analyses)

Eleven of the 15 studies in the meta-analyses wereconducted in the United States,18,21,22,25,27,31–33,36,37,40 two inBelgium,19,39 one in the United States and Canada,35 andone in France.38 They were generally medium to large insize (median sample size, 674; range, 10538–5,04237).Four studies compared tailored with generic nutritioneducation18,27,33,36 and nine compared tailored nutritioneducation with no nutrition education.21,22,32,35,37–40,43 Twostudies included both generic and no nutrition educationcomparison groups.19,31

Study populations included in the meta-analysesusually consisted of healthy or mixed-health-status vol-unteers although they also included the overweight orobese,36 cancer patients,18 and diabetics.28,38 Age ranged

from 18 to �85 years. Most trials included predominantlywhite, female participants, although four included highproportions of priority ethnic groups.21,22,26,29 Two studiesincluded males only,35,37 four included approximatelyequal numbers of males and females,18,28,30,38 and theremainder comprised predominantly female participants.

Priority ethnic and low-income groups(narrative summaries)

The four trials that included high numbers of priorityethnic groups were all conducted in the United Statesand ranged in size from 357 to 3,737 participants (totaln = 5,981).21,22,26,29 Three compared tailored nutritioneducation with no nutrition education21,22,29 and one com-pared tailored nutrition education with generic nutrition

127 studies identified via search

strategy

35 studies obtained for further

review

93 studies excluded

• 52 no print, email or

intervention format excluding

third person

• 19 duration < 6 months

• 12 not parallel design or RCT

• 4 outcomes unsuitable

• 4 not tailored as per definition

• 1 participants < 18yrs

25 studies met the inclusion

criteria

10 studies excluded:

• 2 no print, email or intervention

format excluding face-to-face

• 4 outcomes unsuitable

• 1 duration < 6 months

• 1 not tailored as per definition

• 2 data unsuitable

16 studies included:

• 15 in meta-analyses of all population

groups

• 5* in priority ethnic narrative sub-group summary

• 1* in low-income sub-group narrative summary

(*4/5 trials in sub-group analyses were

also included in the meta-analyses of all

population groups)

9 studies excluded from meta-

and sub-group analyses:

• Did not measure daily servings

of fruits and vegetables or

percentage of energy from total

fat, or data were not suitable for

inclusion in meta-analyses and

authors could not be contacted

• Did not focus on a priority

ethnic or low-income population

Figure 1 Identification and selection of studies.

Nutrition Reviews® Vol. 67(8):464–480 467

Tabl

e2

Char

acte

rist

ics

and

resu

lts

ofst

udie

sw

ith

outc

omes

excl

uded

from

met

a-an

alys

es.

Firs

tAut

hor

(Yea

r),

Coun

try

Stud

ypo

pula

tion

Gro

ups

Follo

w-u

pAd

ditio

nald

ieta

ryan

dan

thro

pom

etric

outc

ome(

s)*

Resu

ltsO

utco

me

mea

sure

men

tin

stru

men

tsO

vera

llris

kof

bias

scor

e

Ande

rson

(200

1),

USA

2529

6M

&F

(96%

),40

y(1

9–77

y),

92%

whi

te,a

nyhe

alth

stat

us,

supe

rmar

ketr

ecru

itmen

t

Tailo

red

(n=

148)

vs.

cont

rol(

n=

148)

6m

onth

sFi

beri

ntak

e(g

/100

0kc

al)

+Sh

oppi

ngre

ceip

ts(li

nked

toa

nutr

ient

data

base

);Bl

ock

FFQ

62

Med

ium

Frui

tand

vege

tabl

ein

take

(g/1

000

kcal

daily

)+

Fibe

rpur

chas

es0

Frui

tand

vege

tabl

epu

rcha

ses

0

Blal

ock

(200

2),

USA

2471

4F,

47y

(40–

56y)

,96.

5%w

hite

,he

alth

yan

dre

cent

lyex

perie

nced

orap

proa

chin

gm

enop

ause

,co

mm

unity

recr

uitm

ent(

clus

ter

desi

gn)

Gen

eric

(n=

105)

,ta

ilore

d(n

=11

4),

gene

ric+

com

mun

ityin

terv

entio

n(n

=16

9),

tailo

red

+co

mm

unity

inte

rven

tion

(n=

159)

6m

onth

sCa

lciu

min

take

(mg/

day)

byst

age

ofch

ange

mod

el44

0(u

neng

aged

)Ab

brev

iate

dBl

ock

NCI

HH

HQ

63M

ediu

m+

(eng

aged

)-

(act

ion)

12m

onth

sCa

lciu

min

take

(mg/

day)

byst

age

ofch

ange

mod

el44

0(u

neng

aged

and

enga

ged)

-(a

ctio

n)

Byrn

e(2

006)

,Au

stra

lia20

74M

&F

(%no

trep

orte

d),3

7.6

y(in

terv

entio

n)an

d38

.6y

(usu

alca

re),

ethn

icity

notr

epor

ted,

over

wei

ghta

ndob

ese,

sede

ntar

y,w

eigh

tsta

ble,

“rea

dyto

chan

ge”

acco

rdin

gto

TTM

,44co

mm

unity

recr

uitm

ent

Tailo

red

(n=

33)a

ndge

neric

(n=

41)

8m

onth

sW

eigh

t(kg

)+

Dig

itals

cale

,st

adio

met

er,

mea

surin

gta

pean

dD

EXA

(dua

len

ergy

x-ra

yab

sorp

tiom

etry

scan

Med

ium

Wai

stci

rcum

fere

nce

(cm

)+

Fatm

ass

(kg)

+

Cam

pbel

l(19

99),

USA

213,

737

M&

F(7

3%),

53.8

y,98

%Af

rican

Amer

ican

,any

heal

thst

atus

,rec

ruitm

entn

otre

port

ed.

Tailo

red

and

cont

rol

(nno

trep

orte

d;pe

rgr

oup

~1,8

69)

24m

onth

sFr

uiti

ntak

e(s

ervi

ngs/

day)

+N

CI15

-item

FFQ

64M

ediu

m

Vege

tabl

ein

take

(ser

ving

s/da

y)+

Cam

pbel

l(20

02),

USA

2266

0F,

mea

nag

eno

trep

orte

d,53

%�

40y

(18–

50y)

,58%

Afric

anAm

eric

an;r

estw

hite

and

othe

r,an

yhe

alth

stat

us,b

lue

colla

rw

orke

rsin

rura

lcou

ntie

s,w

orks

itere

crui

tmen

t(cl

uste

rdes

ign)

Tailo

red

(n=

362)

and

cont

rol(

n=

298)

6m

onth

sFr

uiti

ntak

e+

Brie

f28-

item

food

freq

uenc

ych

eckl

ist

base

don

Bloc

kFF

Q62

Med

ium

Vege

tabl

ein

take

(ser

ving

s/da

y)-

Tota

lfat

(g/d

ay)

018

mon

ths

Frui

tint

ake

0Ve

geta

ble

inta

ke(s

ervi

ngs/

day)

0To

talf

at(g

/day

)0

Clut

terS

nyde

r(2

007)

,USA

2318

2M

&F

(%no

trep

orte

d),m

ean

age

notr

epor

ted

(�65

y),8

2%w

hite

,with

in18

mth

sdi

agno

sis

ofbr

east

orpr

osta

teca

ncer

,se

dent

ary

tom

oder

atel

yac

tive,

canc

erre

gist

ry,p

hysi

cian

and

self-

refe

rral

recr

uitm

ent

Tailo

red

(n=

89)a

ndco

ntro

l(n

=93

)6

mon

ths

BMI

0Th

ree

24-h

rdie

trec

alls

(bas

elin

e,6

mon

ths

and

12m

onth

s)

Med

ium

12m

onth

sBM

I0

De

Bour

deau

dhui

j(2

007)

,Bel

gium

1953

9M

&F

(%no

trep

orte

d),a

geno

tre

port

ed,e

thni

city

notr

epor

ted,

any

heal

thst

atus

,rec

ruitm

entb

ym

edia

tion

betw

een

heal

thpr

omot

ion

serv

ices

and

wor

ksite

s)(c

lust

erde

sign

)

Tailo

red

(n=

192)

,ge

neric

(n=

197)

and

cont

rol(

n=

150)

6m

onth

sov

ertim

eEn

ergy

from

fat(

%)

+(t

ailo

red

vs.

gene

rican

dta

ilore

dvs

.con

trol

)

Bloc

kD

H65

Hig

h

Tota

lfat

inta

ke(g

/day

)+

(tai

lore

dvs

.ge

neric

and

tailo

red

vs.c

ontr

ol

Nutrition Reviews® Vol. 67(8):464–480468

Dem

ark

Wah

nefr

ied

(200

7),

USA

18

543

M&

F(4

6%),

~57

y(2

2–85

y),

83%

whi

te,r

estb

lack

and

othe

r,ea

rlyst

age

brea

stan

dpr

osta

teca

ncer

patie

nts

,can

cerr

egis

try

and

onco

logy

prac

tice

recr

uitm

ent

Tailo

red

(n=

271)

and

gene

ric(n

=27

2)12

mon

ths

%ca

lorie

sfr

omsa

tura

ted

fat

+Bl

ock

NCI

FFQ

66Lo

w

BMI(

kg/m

2 )+

Elde

r(20

06),

USA

2635

7Sp

anis

h-sp

eaki

ngLa

tino

hous

ehol

ds(m

ean

size

~4.7

peop

le),

39.7

y,al

lHis

pani

c/La

tino

(with

wom

an18

–65

yin

hous

ehol

d),a

nyhe

alth

stat

us,

rand

omdi

gitd

ialin

gre

crui

tmen

tus

ing

His

pani

csu

rnam

es

Tailo

red

(118

),ge

neric

(n=

119)

,per

sona

lized

coun

selin

g+

tailo

red

(prin

t)(n

=12

0)

6m

onth

sTo

talf

atin

take

(g/d

ay)

0(a

llco

mpa

rison

s)Th

ree

cons

ecut

ive

24-h

diet

reca

llsat

base

line,

6m

onth

s,an

d12

mon

ths

Med

ium

Satu

rate

dfa

tint

ake

(g/d

ay)

0(a

llco

mpa

rison

s)

12m

onth

sTo

talf

atin

take

(g/d

ay)

0(a

llco

mpa

rison

s)

Satu

rate

dfa

tint

ake

(g/d

ay)

0(a

llco

mpa

rison

s)

Jone

s(2

003)

,Ca

nada

2810

29M

&F

(47.

6%),

~54

y(r

ange

notr

epor

ted)

,eth

nici

tyno

tre

port

ed(a

ble

tore

adan

dun

ders

tand

Engl

ish)

,ove

rwei

ght,

high

-fat

diet

,dia

betic

,loc

alfa

mily

prac

tice,

diab

etes

educ

atio

nce

nter

s,Ca

nadi

anD

iabe

tes

Asso

ciat

ion,

and

com

mun

ityre

crui

tmen

t

Tailo

red

(n=

510)

and

gene

ric(n

=51

9)12

mon

ths

Frui

tint

ake

(ser

ving

s/da

y)+

NCI

Bloc

kFF

Q67

Low

Vege

tabl

ein

take

(ser

ving

s/da

y)+

Lutz

(199

9),U

SA31

710

M&

F(6

4.4%

),39

.3y,

77.9

%w

hite

,res

tbla

ck,a

nyhe

alth

stat

us,s

ubsc

riber

sto

ahe

alth

mai

nten

ance

orga

niza

tion

inN

orth

Caro

lina

(rec

ruitm

ent)

Tailo

red

(n=

176)

,ta

ilore

d+

goal

sett

ing

(n=

177)

,gen

eric

(n=

177)

,and

cont

rol

(n=

180)

6m

onth

sVa

riety

offr

uita

ndve

geta

bles

cons

umed

(num

ber/

wee

k)+

(bot

hta

ilore

dgr

oups

vs.c

ontr

ol)

17-it

emFF

Qde

velo

ped

from

thre

eFF

Qs

from

sim

ilar

stud

ies

Low

0(b

oth

tailo

red

grou

psvs

.gen

eric

and

tailo

red

vs.

tailo

red

+go

alse

ttin

g)

Nitz

ke(2

007)

,U

SA33

2042

M&

F(6

1.2%

),m

ean

age

not

repo

rted

(18–

24y)

,53.

7%w

hite

,27

.1%

Afric

anAm

eric

an,r

est

othe

r,an

yhe

alth

stat

us,y

oung

,ec

onom

ical

lydi

sadv

anta

ged

adul

ts,p

erso

nalc

onta

cts

and

adve

rtis

emen

tsin

loca

tions

with

high

prop

ortio

nof

low

-inco

me

(rec

ruitm

ent)

Tailo

red

and

gene

ric(n

umbe

rspe

rgro

upno

trep

orte

dn~

1021

)

12m

onth

sVe

geta

ble

inta

ke(s

ervi

ngs/

day)

0FF

Q(5

+a

day

ques

tionn

aire

)M

ediu

m

Frui

tint

ake

(ser

ving

s/da

y)+

Perc

eive

dda

ilyin

take

ques

tion

Roth

ert(

2006

),U

SA34

2862

M&

F(8

3%),

~45

y,~5

6%w

hite

,36%

Afric

anAm

eric

an,r

est

His

pani

can

dot

her,

over

wei

ght/

obes

e,m

embe

rsof

non-

profi

tin

tegr

ated

heal

thca

resy

stem

inG

eorg

ia,M

id-A

tlant

icSt

ates

,N

orth

wes

tand

Ohi

o(r

ecru

itmen

t)

Tailo

red

(n=

1475

)and

gene

ric(n

=13

87)

6m

onth

sPe

rcen

tage

ofbo

dyw

eigh

tlos

t+

Self-

repo

rtw

eb-b

ased

ques

tionn

aire

Med

ium

Tota

lbod

yw

eigh

tlos

t(kg

)+

Nutrition Reviews® Vol. 67(8):464–480 469

Tabl

e2

Co

nti

nu

edFi

rstA

utho

r(Y

ear)

,Co

untr

y

Stud

ypo

pula

tion

Gro

ups

Follo

w-u

pAd

ditio

nald

ieta

ryan

dan

thro

pom

etric

outc

ome(

s)*

Resu

ltsO

utco

me

mea

sure

men

tin

stru

men

ts

Ove

rall

risk

ofbi

assc

ore

Tate

(200

7),U

SA36

192

M&

F(8

4%),

19.2

y(2

0–65

y),

prio

rity

popu

latio

n<1

0%,a

nyhe

alth

stat

us,o

verw

eigh

tand

obes

e,re

crui

tmen

tnot

repo

rted

.

Tailo

red

+au

tom

ated

tailo

red

e-m

ail

feed

back

(n=

61),

tailo

red

e+

e-m

ail

coun

selin

gfe

edba

ckfr

omdi

etiti

an(n

=64

),an

dge

neric

(n=

67)

6m

onth

sCh

ange

inbo

dyw

eigh

t(kg

)+

(tai

lore

d+

diet

itian

feed

back

vs.g

ener

ic)

Bloc

kFF

Q63

Med

ium

0(t

ailo

red

+e-

mai

lfee

dbac

kvs

.gen

eric

)Pe

rcen

tage

ofin

itial

body

wei

ghtl

ost(

%)

+(t

ailo

red

+di

etiti

anfe

edba

ckvs

.gen

eric

)0

(tai

lore

d+

e-m

ailf

eedb

ack

vs.g

ener

ic)

Perc

enta

geof

ener

gyfr

omto

talf

at(%

)+

(tai

lore

d+

diet

itian

feed

back

vs.g

ener

ic)

0(t

ailo

red

+e-

mai

lfee

dbac

kvs

.gen

eric

)

Tille

y(1

999)

,USA

3750

42M

&F

(7%

),55

y(in

terv

entio

n)an

d58

y(c

ontr

ol),

93%

whi

te,a

nyhe

alth

stat

us,

wor

kers

inth

eau

tom

otiv

ein

dust

ry,w

orks

ites

alre

ady

enro

lled

ina

colo

rect

alsc

reen

ing

prog

ram

(rec

ruitm

ent;

clus

ter

desi

gn)

Tailo

red

(n=

2240

)and

cont

rol(

n=

2802

)12

mon

ths

Fibe

rint

ake

(g/1

000

kcal

)+

Mod

ified

NCI

FFQ

63H

igh

24m

onth

sFi

beri

ntak

e(g

/100

0kc

al)

+

Turn

in(1

992)

,Fr

ance

3810

5M

&F

(43%

),44

.8y,

ethn

icity

notr

epor

ted,

any

heal

thst

atus

,w

ithty

peon

eor

type

two

diab

etes

,rec

ruitm

entm

etho

dno

tre

port

ed.

Tailo

red

(n=

54)a

ndco

ntro

l(n

=51

)6

mon

ths

Calo

ricex

cess

(kca

l)+

Chan

gein

eatin

gha

bits

(thr

ough

3-da

ydi

etar

yre

cord

san

din

vest

igat

ion

atpa

rtic

ipan

tsho

me

bydi

etiti

an)

Med

ium

Carb

ohyd

rate

defic

it(%

)+

Fati

ntak

eex

cess

(%)

+Pe

rcen

tage

ofen

ergy

from

carb

ohyd

rate

%0

Nutrition Reviews® Vol. 67(8):464–480470

Vand

elan

otte

(200

5),B

elgi

um39

1023

M&

F(6

4.5%

),39

.1y

(20–

60y)

,eth

nici

tyno

trep

orte

d,he

alth

y,co

mm

unity

recr

uitm

ent

(pos

ters

,lea

flets

,and

e-m

ail)

Tailo

red

PAan

dfa

tin

take

(sim

ulta

neou

s),

tailo

red

PAat

base

line

and

fati

ntak

eat

3m

onth

san

dta

ilore

dfa

tint

ake

atba

selin

ean

dPA

at3

mon

ths

(seq

uent

ial),

cont

rol.

nno

trep

orte

dfo

rea

chgr

oup

~341

6m

onth

sTo

talf

atin

take

(g/d

ay)

+(b

oth

tailo

red

grou

pslo

wer

inta

keth

anco

ntro

l)

FFQ

(48

item

,va

lidat

ed)65

;not

e:%

ener

gyfr

omto

tal

fate

stim

ated

usin

gre

com

men

ded

ener

gyin

take

tabl

es,a

sFF

Qdi

dno

tmea

sure

ener

gy.

Med

ium

Perc

enta

geen

ergy

from

satu

rate

dfa

t(%

)+

(bot

hta

ilore

dgr

oups

low

erin

take

than

cont

rol)

Win

nett

(199

7),

USA

4012

7M

&F

(86%

),40

y(1

9–77

y),

95%

whi

te,a

nyhe

alth

stat

us,

prim

ary

shop

pers

,sup

erm

arke

tflo

orre

crui

tmen

t

Tailo

red

(n=

54)a

ndco

ntro

l(n

=51

)6

mon

ths

Perc

enta

geca

lorie

sfr

omto

tal

fati

nfiv

efo

odgr

oups

(%)

+(d

airy

prod

ucts

and

allf

oods

–da

iry)

Mea

sure

dby

supe

rmar

ketr

ecei

pts

linke

dto

food

nutr

ient

data

base

Hig

h

Fibe

r(g/

1000

cal)

inth

ree

food

grou

ps+

Frui

tand

vege

tabl

ese

rvin

gs(s

ervi

ngs/

1000

cal)

0

Win

nett

(200

7),

USA

4110

71M

&F

(76%

),53

y,23

%Af

rican

Amer

ican

,oth

eret

hnic

ityno

trep

orte

d,he

alth

ych

urch

mem

bers

,rec

ruite

dfr

omch

urch

esin

Virg

inia

(clu

ster

desi

gn)

GTH

tailo

red

inte

rven

tion

(n=

364)

,G

THta

ilore

din

terv

entio

nw

ithch

urch

-bas

edsu

ppor

ts(G

+)(n

=36

4),a

ndco

ntro

l(n

=34

3)

6m

onth

sFi

beri

ntak

e(g

/100

0kc

al)

+(G

THta

ilore

dvs

.co

ntro

l;0

foro

ther

com

paris

ons)

FFQ

68co

mbi

ned

with

groc

ery

rece

ipts

(5w

eeks

ofin

form

atio

nlin

ked

toa

nutr

ient

data

base

)

Med

ium

Frui

tand

vege

tabl

ein

take

(ser

ving

s/10

00kc

al)

+(G

THan

dG

TH+

vs.c

ontr

ol)

Wyl

ie-R

oset

t(2

001)

,USA

4258

8M

&F

(82.

3%),

52.2

y,83

%w

hite

,ove

rwei

ghto

robe

sew

ithCV

Dris

kfa

ctor

(s),

com

mun

ityad

vert

isin

gan

dm

embe

rsof

ahe

alth

mai

nten

ance

orga

niza

tion

(rec

ruitm

ent)

Tailo

red

(GT)

(n=

236)

,ta

ilore

d+

cons

ulta

tion

(GTC

)(n

=23

6),a

ndge

neric

(G)(

n=

116)

12m

onth

sEn

ergy

inta

ke(K

cal/d

ayan

d%

Kcal

from

tota

lfat

)0

(all

com

paris

ons)

Bloc

kFF

Q63

Med

ium

Body

wei

ght(

lb)

+(t

ailo

red

with

cons

ultv

s.ge

neric

;0

foro

ther

com

paris

ons)

BMI(

kg/m

2 )0

(all

com

paris

ons)

Wai

stci

rcum

fere

nce

(inch

es)

0(a

llco

mpa

rison

s)Pe

rcen

tage

body

fat(

%)

0(a

llco

mpa

rison

s)Pe

rcen

tage

wei

ghtl

ost(

%)

0(a

llco

mpa

rison

s)*O

nly

prim

ary

outc

omes

incl

uded

.Ifp

rimar

you

tcom

esno

tdes

crib

ed,a

llou

tcom

esin

clud

ed.

Abbr

evia

tions

:+,i

nfa

voro

ftai

lore

dnu

triti

oned

ucat

ion;

0,no

sign

ifica

ntdi

ffere

nce

betw

een

grou

ps;-

,in

favo

rofg

ener

icnu

triti

oned

ucat

ion

orco

ntro

l;D

H,d

ieth

isto

ry;F

,fem

ale;

FFQ

,foo

dfr

eque

ncy

ques

tionn

aire

;GTH

,gui

deto

heal

th;H

HH

Q,h

ealth

hist

ory

habi

tsqu

estio

nnai

re;M

,mal

e;N

CI,N

atio

nalC

ance

rIns

titut

e;PA

,phy

sica

lact

ivity

.

Nutrition Reviews® Vol. 67(8):464–480 471

education.26 Participants were predominantly femaleand African American,21,22,29 although in one trial theywere 100% Hispanic/Latino.26 Age ranged from 18 to65 years.

The single trial of a low-income group was large(n = 2,042) and completed in the United States. Nitzkeet al.33 compared tailored with generic nutrition educa-tion in a predominantly female population (~61%) ofyoung adults (18–24 years).

INTERVENTIONS AND OUTCOMESOF INCLUDED STUDIES

All population groups (meta-analyses)

Most interventions were tailored by current diet, nutrientintake, or food purchases, as well as components ofbehavior theories (the most common being the trans-theoretical model by Prochaska et al.44). Other tailoringframeworks included the following: nutrition knowledge;perceived adequacy of nutrient intake; dietary prefer-ences; occupation; anthropometry; demographics; healthconcerns and behaviors; diabetic profile; environmentand social support; religion; motivational reasons forwanting to lose weight or change diet; former weight-lossexperiences; and psychosocial factors. Most tailoredinterventions were delivered by newsletters, pamphlets,magazines, or workbooks. However, e-mail and theInternet were also common, and two studies deliverededucation via a kiosk housed in a supermarket.25,40

The frequency of tailored feedback ranged from one to36 occasions over intervention periods of 1 day to18 months.

Priority ethnic and low-income groups(narrative summaries)

Trials of priority ethnic groups (n = 4) used very similarframeworks for tailoring, including use of the transtheo-retical model (TTM).44 However, all four studies deliv-ered tailored nutrition education by print materials.Frequency of feedback ranged from two to 12 occasionsover intervention periods of 12 weeks to 20 months. Out-comes assessed included fruit and vegetable intake (threetrials), fruit intake alone and vegetable intake alone(n = 2), total fat intake (n = 2), and saturated fat intake(n = 1).

The single trial involving a population group on alow-income33 tailored the nutrition education withcurrent fruit and vegetable intake and components of theTTM.44 Eight packages of print mailers and two phonecalls were delivered over a 4-month intervention period.33

Outcomes assessed were as follows: combined fruit andvegetable intake; fruit intake alone; and vegetable intakealone.33

QUALITY OF INCLUDED STUDIES

Overall, the quality of the studies was moderate to good.Quality indicators for all trials are described below:

All population groups (meta-analyses)

All 15 studies were randomized or quasi-randomizedcontrolled trials. Only three reported how their randomallocation sequence was generated18,36,37 and two reportedconcealment of their random allocation sequence.18,33

Two trials did not report baseline characteristics, nor didthey report them by treatment group.32,45 Of the remain-ing 13 trials, six did not have equivalent groups at base-line19,21,22,31,37,41; only one failed to adjust for potentialconfounders in their analysis.19 Loss to follow-up wasreported for all 15 trials and ranged from 1% to 43%.19,35

However, losses by study group were not reported forseven trials.21,25,27,31,33,39,40 Four of the eight trials withgroup losses had losses of �5% between groups.19,22,36,38

Blinding of study outcome assessors was reported foronly two trials.18,36 Similarly, only three trials reported ana priori sample size calculation.18,36,41 Notably, only five(33%) trials included intent-to-treat analyses.18,19,33,35,37,46

Three trials in the meta-analyses received an overall“high” risk of bias,22,37,40 two a “low” risk of bias,18,31 andthe remaining 10 trials received an overall “medium” riskof bias.19,21–23,25,27,32,33,35,36,38,39

Priority ethnic and low-income groups(narrative summaries)

All four trials including priority ethnic groups were ran-domized controlled trials. Only one reported how therandom allocation sequence was generated and that theallocation was adequately concealed.29 Three of the fourstudies had equal groups at baseline.21,26,29 However, theremaining trial adjusted for baseline differences.22 Lossto follow-up ranged from 18% to 33%21,22 and, wherereported, was equal across groups. Blinding of studyassessors and a priori sample size were not reported byany study. Three of the four priority ethnic trials weredetermined to have an overall “medium” risk of bias,21,26,29

and one a “high” risk of bias.22

The single trial assessing a low-income populationwas a randomized controlled trial. Nitzke et al.33 reportedadequate random allocation and concealment methodsand had equal groups at baseline. However, lossto follow-up was high (approximately 39%) and not

Nutrition Reviews® Vol. 67(8):464–480472

reported according to group. Further, sample size calcu-lation was not reported. This trial received an overall“medium” risk of bias.33

Publication bias

A funnel plot was created using the outcome measured bythe greatest number of trials in the review (fruit andvegetable intake in servings/day; n = 9). Funnel plots aresimple scatter plots of the treatment effects estimatedfrom the individual studies (on the x-axis) versus eachstudy’s sample size (on the y-axis),16 and are used toinvestigate whether “the likelihood of finding studies isrelated to the results of those studies” (i.e., publicationbias).47–49 Precision in the estimation of true treatmenteffect increases as sample size increases. Therefore, in theabsence of bias, the scatter plot should resemble aninverted funnel.16 Asymmetry may indicate publicationbias. The funnel plot of studies included in this review(Figure 2) shows a lack of studies across the base (on bothsides of the triangle), indicating that small trials (bothnegative and positive) may be missing.

LONG-TERM EFFECTS OF TAILORING ON DAILYSERVINGS OF FRUITS AND VEGETABLES

Tailored versus generic nutrition education

Four trials (n = 4,638) measured the effect of interven-tions on daily servings of fruits and vegetables and com-

pared tailored nutrition education with generic nutritioneducation.18,27,31,33 Three used a food frequency question-naire (FFQ),27,31,33 and one a dietary history method18 asmeasurement instruments. One included a follow-up at6 months,31 and three at 12 months.18,27,33

The weighted mean difference (WMD) in self-reported intakes between groups following interventionwas 0.35 servings per day (95% CI 0.19–0.52; P < 0.0001)in favor of tailored nutrition education (Figure 3). A fixedeffects model was used as there was no significant hetero-geneity (I2 = 7%, P = 0.36).

Tailored nutrition education versus no nutritioneducation (control)

Six trials (n = 12,187) measured the effect of interven-tions on daily servings of fruits and vegetables and com-pared tailored nutrition education with no nutritioneducation (control).21,22,31,32,35,37 All six used an FFQ asthe measurement instrument.21,22,31,32,35,37 The longestfollow-up times were 6 months (3 trials),21,31,35 12 months(1 trial),32 and �12 months (2 trials).22,37

Following intervention, the WMD in self-reportedintakes (0.59 servings per day; 95% CI, 0.21–0.98;P = 0.002) was significantly in favor of tailored nutritioneducation (Figure 4). Due to significant heterogeneity(I2 = 88.9%; P < 0.00001) a random effects model wasused. Two trials had a high risk of bias (due to failure toreport blinding, random and concealed allocationmethods and failure unequal groups at baseline)22,37; thus,

Figure 2 Funnel plot of nine trials reporting daily servings of fruits and vegetables.

Figure 3 Meta-analysis of trials comparing tailored and generic education with daily servings of fruits and vegetablesas the outcome (fixed effects model).

Nutrition Reviews® Vol. 67(8):464–480 473

a sensitivity analysis was completed (data not shown).When the two studies with a high risk of bias wereremoved from the analysis, the WMD between groupsincreased somewhat and the confidence remained ofsimilar width (0.81 servings per day; 95% CI, 0.42–1.21;P < 0.0001).

LONG-TERM EFFECTS OF TAILORING ON PERCENTAGEOF ENERGY FROM TOTAL FAT

Tailored versus generic nutrition education

Three trials (n = 1,060) measured percentage of energyfrom total fat (%ETF) and compared tailored nutritioneducation with generic nutrition education.18,19,36 Two ofthe three studies used a diet history18,19 and one an FFQto measure fat intake.36 Two of the trials included afollow-up at 6 months19,36 and the remaining trial at 12months.18

Following intervention, the WMD in self-reportedintakes was -2.2 %ETF, (95%CI -3.0–-1.4; P < 0.00001)(Figure 5). A fixed effects model was used as there was nosignificant heterogeneity (I2 = 0%; P = 0.58).

Tailored nutrition education versus no nutritioneducation (control)

Six studies (n = 6,572) that measured percentage ofenergy from total fat (%ETF) compared tailored nutritioneducation with no nutrition education.19,25,37–40 Two used

an FFQ,37,39 one used a dietary history method,19 one a3-day diet record,38 one supermarket sales receipts,40 andthe remaining study used both supermarket sales receiptsand an FFQ25 as measurement instruments. Five trialsincluded a follow-up at 6 months19,25,38–40 and the remain-ing trial at �12 months.37

Following intervention, the WMD between groupswas -2.45%ETF (95% CI -4.08–-0.82; P = 0.0005) infavor of tailored nutrition education (n = 3,029)(Figure 6). A random effects model was used to drive thesummary statistic as there was significant heterogeneity(I2 = 80%; P = 0.0003).

LONG-TERM EFFECTIVENESS OF TAILORED NUTRITIONEDUCATION FOR PRIORITY ETHNIC AND

LOW-INCOME GROUPS

Priority ethnic groups

Of the four trials that focused on priority ethnicgroups,21,22,26,29 three used an FFQ as the measurementinstrument,21,22,29 and the remaining study used three24-h diet recalls.26

For daily servings of fruits and vegetables, resultswere consistently in favor of tailored nutrition education(looking at the longest follow-up points): Campbell et al.21

reported a mean (SE) difference of 0.85 (0.12) servingsper day between tailored and control groups at 24months. At 18 months, Campbell et al.22 reported a sig-nificantly higher (P < 0.05) fruit and vegetable intake in

Figure 4 Meta-analysis of trials comparing tailored nutrition education and control, with daily servings of fruits andvegetables as the outcome (random effects model).

Figure 5 Meta-analysis of trials comparing tailored and generic nutrition education, with percentage of energy fromtotal fat as the outcome (fixed effects model).

Nutrition Reviews® Vol. 67(8):464–480474

participants that received tailored nutrition [mean (SD),3.6 (3.1) servings/day] compared with those that receivedno nutrition education [mean (SD), 3.4 (2.9) servings/day], and Krueter et al.29 reported a significantly(P < 0.03) higher intake in the tailored group (mean dif-ference +0.96 servings/day) compared with the controlgroup (mean difference 0.25 servings/day).

None of the priority ethnic studies assessed theefficacy of tailored nutrition education to decrease thepercentage of energy from total fat. However, two tri-als22,26 measured total fat intake (g). Neither of thesestudies found a significant difference between groups(at 18 and 12 months, respectively).22,26 In contrast, twostudies assessed fruit intake and both reported signifi-cantly higher servings per day in the tailored groupcompared with the control group [mean difference at24 months (SE), 0.66 (0.09) servings/day (P = 0.001)21;mean (SD) tailored, 1.9 (2.0) and mean (SD) control,1.7 (1.9) (P = 0.02) at 18 months].22 Similarly, the twostudies that assessed vegetable intake alone reportedresults in favor of tailored nutrition education: Camp-bell et al.21 found a significantly (P = 0.00003) higherincrease in servings of vegetables per day in the tailoredgroup compared with the control group at 24 months[mean (SE) difference, 0.19 (0.04)]. Campbell et al.22

reported an unadjusted mean difference of -0.2 servingsper day (P = 0.03, based on time by group interaction) at18 months. One trial26 compared saturated fat intake(g/day) between groups that received tailored or genericnutrition education, finding no significant difference at12 months.

Low-income

One study specifically included participants on a lowincome (defined as participating in a public assistanceprogram or self-reported income <US$16,000).33 An FFQand a perceived daily intake question were used as mea-surement instruments.33

Using an FFQ, Nitzke et al.33 found at 12 monthssignificantly higher self-reported servings of fruit alone

and fruit and vegetables combined in the tailored group,compared with the control group [fruit intake mean (SD),2.5 (2.2) servings/day in the tailored group and 2.2 (2.2)servings/day in the control group (P < 0.05); vegetableintake mean (SD), 4.3 (2.9) servings/day in the tailoredgroup and 3.9 (3.1) servings per day in the control group(P < 0.05)]. There were no differences between groups forservings of vegetables alone (P > 0.05).33 Using the per-ceived intake question, self-reported intake of fruits andvegetables combined, fruit alone, and vegetables alonewere significantly higher in the tailored group comparedwith the control group (at 12 months).33

DISCUSSION OF FINDINGS

Based on the evidence to date, tailored nutrition educa-tion appears to be a promising strategy for improving thedietary intake of adults over the long term (6 months orlonger), including for priority ethnic and low-incomepopulation groups. The majority of studies have focusedon fruit and vegetable intake in servings/day, and per-centage of energy consumed from total fat. Our meta-analyses suggest tailored nutrition education was moreeffective than generic nutrition education and control:Those in the tailored group were found to consumeWMD 0.35 servings of fruits and vegetables per day moreand WMD –2.20% ETF less than participants receivinggeneric nutrition education. In addition, participants thatreceived tailored nutrition education were found toconsume WMD 0.59 servings of fruits and vegetables perday more and WMD -2.45 %ETF less than participantsthat received no nutrition education. Few studies of tai-lored nutrition education included large numbers of par-ticipants from a priority ethnic (n = 4) or low-income(n = 1) group. However, those that did were large (range,357 to 3,737 participants), and most reported a positiveeffect of tailored nutrition education compared withcontrol.

This review was conducted according to a protocolin line with that described in the Cochrane Handbook for

Figure 6 Meta-analysis of trials comparing tailored nutrition education and control, with percentage of energy fromtotal fat as the outcome (fixed effects model).

Nutrition Reviews® Vol. 67(8):464–480 475

Systematic Reviews of Interventions.16 As such, the meth-odology was rigorous, bias and chance effects werelimited, and the results are likely to be reliable. Further,the use of meta-analyses to summarize the effects for alladults provides more precise estimates than narrativesummary.16

However, trials included in the review were subjectto several methodological limitations: many reportedmultiple outcomes and did not specify which (if any)outcome was primary. Including multiple outcomes orendpoints in a study may increase the overall rate of typeI error (false positive), because several significant resultscan be expected to occur by chance alone. For example: ifthe level of significance (a) is set at 0.05 (5%) for oneoutcome, then in a study such as that by Wylie Rosettet al.42 which included 13 different outcomes, the pro-bability of finding at least one significant result is1 - (1-0.05)13 = 0.49 (49%).50 Six studies in this reviewincluded more than three primary outcomes (all out-comes were included if the primary was not specified).Therefore, it is possible that some of the positive effectsreported were in fact the result of chance, rather than trueefficacy of the tailored nutrition education.

Another important limitation was that most studies(13 of 15 in the meta-analyses, and five of five in thesubgroup summaries) used self-reported outcome mea-sures, such as diet records, diet histories, and FFQs.Therefore, it is possible that social desirability andmemory bias caused some participants to report havingmade recommended dietary changes even if they did not.These biases are less of an issue when all participants in atrial express the same need to provide socially desirableinformation or have the same memory skills51 (such aswhen tailored and generic nutrition education are com-pared). However, when these biases are likely to beuneven between groups (such as when tailored nutritioneducation is compared with control), it is possible that thevalidity of dietary data is compromised and the resultsconsequently biased toward tailored nutrition education.Two trials in the review compared tailored nutritioneducation with a control and used objective measureof dietary intake: one used solely an objective measure(shopping receipts),40 while the other used shoppingreceipts in addition to data from an FFQ.25 Both studiesreported significant differences in favor of tailored nutri-tion education. However, when Anderson et al.25 assesseddietary data obtained from shopping receipts separatefrom data obtained using FFQs, they found significanteffects of tailored education only with the self-reportedFFQ data. These results suggest it is possible that some ofthe effects of tailoring observed in the review were due toself-reporting and/or memory bias.

Also, three trials in the meta-analysis carried a highrisk of bias, which may have affected their results.22,37,40

Nevertheless, a sensitivity analysis suggested true efficacyof tailored nutrition education because the summary sta-tistics and confidence intervals changed little when thetrials with high-level bias risks were removed.

In addition to having methodological limitations, thetrials in this review assessed a wide range of dietary out-comes. This made them difficult to combine and, conse-quently, only two outcomes could be appraised in themeta-analyses. Further, although the term “tailoring”implies a technique that is fairly standardized, the tailor-ing frameworks included in this review were diverse. It ispossible that some components of tailoring frameworkswere more or less effective than others. If this was true, theprecision of the review may have been lowered.

Lastly, it is notable that publication bias was sug-gested in this review. The funnel plot indicated a lackof small studies (both positive and negative), and trialswere only included if they were published in English.However, a funnel plot is a low-powered test and thenumber of studies included was small (n = 9). Further,due to the computerized nature of tailored interventions,it is relatively simple for researchers to deliver the inter-ventions to large groups. Thus, the gap in the funnel plotmay be due to an actual lack of small trials in the area,rather than publication bias. Nevertheless, reliance onpublished studies and/or those published in English mayhave led to an overestimation of the effect of tailorednutrition education.52

The findings of this review build upon the results ofanother review in the area. Kroeze et al.15 reviewed ran-domized controlled trials of computer-tailored interven-tions to change physical activity and dietary behaviorsin adults.15 A total of 26 diet-related behavior trials thatwere published in English between 1965 and 2004 wereincluded and six of them reported effects over the longterm (�6 months).15 Methodological and clinical hetero-geneity prevented pooling of data, resulting in a narrativereview.15 The main differences between the two reviewswere the publication dates (the current review onlyincluded trials published after 1990), trial duration (theprevious review included trials �3 months, althoughthese were reviewed separately, and it defined long-termtrials as >6 months), and analyses of sub-groups athighest risk of nutrition-related disease (i.e., priorityethnic and low-income) in the current review.

Similar to the present review, Kroeze et al.15 reportedsignificant long-term (>6 months) effects in favor of tai-lored nutrition education for combined fruit and veg-etable intake (although only two studies were included inthis section). They did not assess %ETF, but they found nodifference between groups for intake of total fat (g; n = 2studies) at >6 months (although the different definitionsof “long-term” (>6 months compared with �6 monthsin the current review) hindered exact comparison of

Nutrition Reviews® Vol. 67(8):464–480476

results). A notable finding from the previous review wasa lack of long-term trials looking at dietary behaviors(n = 4). Six trials from the review by Kroeze et al.15 wereincluded in the current review, indicating that tailorednutrition education is becoming an increasingly populardietary intervention. Narrative reviews of tailored nutri-tion education13,53 have produced similar findings toKroeze et al.,15 and to the current review.

This review has shown that tailored nutrition educa-tion is a promising strategy for improving the dietarybehaviors of all adults over the long term (�6 months).Should automated tailored nutrition education be imple-mented widely as a public health strategy, it is possible thatadults would increase their intake of fruits and vegetablesby approximately 0.6 servings per day, and lower theirintake of percentage of energy from total fat by approxi-mately 2.5%, which are clinically significant improve-ments in light of current intakes and recommendations:according to New Zealand National Nutrition Surveydata,54 38% of adult (16 years and over) males and 27% ofadult females do not consume the recommended �3 serv-ings of vegetables55 per day (figures for Maori and PacificIslanders are 50% and 38%, and 73% and 58%, respec-tively). Similarly for fruits (recommended daily servings�2 per day),55 66% of all adult males and 44% of females donot meet the recommendations (69% and 51% for Maori,and 22% and 12% for Pacific Islanders, respectively). For%ETF, implementing tailored nutrition education couldpotentially decrease intakes for all adults to approximately33% ETF (35% and 33% TE for adult Maori, and 33% and31% for adult Pacific Islander males and females, respec-tively).Recommended intake of total fat in New Zealand is30–35%TE.56 In the United States, recommended intake offruits is two or more servings per day and vegetables, threeor more servings per day (five servings or more of fruitsand vegetables combined per day).57 For total fat, the rec-ommendation is 20–35% of total energy.57 However, datafrom the 1999–2002 National Health and NutritionExamination Surveys found US adults consume onaverage only one serving of fruit and two servings ofvegetables per day (totaling three servings of fruits andvegetables/day),58 and total fat intake is at the higher end ofthe recommended spectrum (33% of energy from totalfat).59 Similar to potential dietary improvements for NewZealand, wide implementation of tailored nutrition edu-cation in the United States could increase fruit and veg-etable intake to 3.6 servings per day and lower percentageof energy from total fat to approximately 30%. There mayalso be positive effects of tailored nutrition education forother important dietary components,1 such as saturatedfat, sodium, and fiber, the recommended intakes of whichmany New Zealanders and Americans do not meet.54,59

Such improvements in dietary quality could have wideimplications for public health.1

Future trials can add to, improve upon, and extendthe evidence included in this review by concentrating ona select set of outcomes most relevant to nutrition-relateddisease (e.g., saturated fat, fiber and energy density, andchange in body weight or BMI). Further, they shoulddetermine sample size a priori based on one primaryoutcome, attempt to use more objective and reliable mea-sures of dietary intake (such as supermarket food pur-chases or biomarkers, in addition to more traditionalmethods of dietary assessment), use appropriate random-ization strategies, adequately conceal allocation, and useintent-to-treat analyses. In addition, it is important thattrial methodology be adequately reported to allow bettercomparison of studies. These recommendations are inline with the Consolidated Standards for Reporting ofTrials (CONSORT).60 Davidson et al.61 recommendedfurther that behavioral medicine researchers “should payparticular attention to the CONSORT items focused onrandomization and blinding procedures, multiplicity ofoutcomes, and the clear identification of which outcomeis primary.”61 Lastly, there is a clear need for further large,high-quality trials, particularly including priority ethnicgroups and low-income groups. Investigation of the useof tailored nutrition education in multiple settings (suchas supermarkets, schools, and workplaces), and for men,would also be beneficial.

CONCLUSION

Research to date suggests tailored nutrition education is apromising strategy for improving the diets of adults overthe long term. However, frequent use of self-reporteddietary data and reporting of multiple outcome measuressuggest a strong possibility of false-positive findings.Future studies should endeavor to use more objectivemeasures of diet, clearly identify the primary studyoutcome, concentrate on outcomes most relevant tonutrition-related disease, and ensure adequate reportingof research design and methods.

Acknowledgments

The authors are grateful to Varsha Parag who providedstatistical advice.

Funding. HE is supported by a National Heart Founda-tion (NHF) of New Zealand post-graduate scholarship(Grant 1285). CNM and HE receive salary support fromthe Health Research Council of New Zealand (Grant06/379)

Declaration of interest. The authors have no competinginterests to declare.

Nutrition Reviews® Vol. 67(8):464–480 477

Authorship responsibilities. HE was responsible for devel-oping the search strategy, conducting the search, choos-ing the included studies, completing the analysis, anddrafting the manuscript. CNM was responsible for pro-viding advice on methodological aspects of the review,the included studies, and revising the manuscript.

References

1. Ezzati M, Lopez AD, Rodgers A, Vander Hoorn S, Murray CJL,and The Comparative Risk Assessment Group. Selected majorrisk factors and global and regional burden of disease.Lancet. 2002;360:1347–1360.

2. James WP, Nelson M, Ralph A, Leather S. Socioeconomicdeterminants of health: the contribution of nutrition toinequalities in health. BMJ. 1997;314:1545–1549.

3. Giskes K, Van Lenthe FJ, Brug J, Mackenbach JP, Turrell G.Socioeconomic inequalities in food purchasing: the contri-bution of respondent-perceived and actual (objectivelymeasured) price and availability of foods. Prev Med. 2007;45:41–48.

4. Ricciuto LE, Tarasuk VS. An examination of income-relateddisparities in the nutritional quality of food selections amongCanadian households from 1986–2001. Soc Sci Med.2007;64:186–198.

5. Turrell G, Hewitt B, Patterson C, Oldenburg B, Gould T. Socio-economic differences in food purchasing behaviour andsuggested implications for diet-related health promotion. JHum Nutr Diet. 2002;15:355–364.

6. Lawes C, Stefanogiannis N, Tobias M, Paki Paki N,Ni Mhurchu C, Turley M, et al. Ethnic disparities in nutrition-related mortality in New Zealand: 1997–2011. N Z Med J.2006;119:U2122.

7. Ministry of Health. The Health of Pacific Peoples. Wellington,New Zealand: Ministry of Health; 2005.

8. Yip YH, Malik I, Luscombe C, McCarry M, Beevers G. Dietary fatpurchasing habits in whites, blacks and Asian peoples inEngland – implications for heart disease prevention. Int JCardiol. 1995;48:287–293.

9. Larsson I, Lissner L, Wilhelmsen L. The “Green Keyhole” revis-ited: nutritional knowledge may influence food selection. EurJ Clin Nutr. 1999;53:776–780.

10. Turrell G, Kavanagh AM. Socio-economic pathways to diet:modelling the association between socio-economic positionand food purchasing behaviour. Pub Health Nutr. 2006;9:375–383.

11. Brug J, van Assama P, Kok G, Lenderink T, Glanz K. Self-rateddietary fat intake: association with objective assessment offat, psychosocial factors, and intention to change. J NutrEduc. 1994;26:218–223.

12. Lechner L, Brug J, De Vries H. Misconceptions of fruit andvegetable consumption: differences between objective andsubjective estimation of intake. J Nutr Educ. 1997;29:313–320.

13. Kreuter MW, Strecher VJ, Glassman B. One size does not fit all:the case for tailoring print materials. Ann Behav Med. 1999;21:276–283.

14. Adamson AJ, Mathers JC. Effecting dietary change. Proc NutrSoc. 2004;63:537–547.

15. Kroeze W, Werkman A, Brug J. A systematic review ofrandomized trials on the effectiveness of computer-tailored

education on physical activity and dietary behaviours. AnnBehav Med. 2006;31:205–223.

16. The Cochrane Library. Section 9. Analysing and presentingresults. In: Higgins J, Green S (eds) Cochrane Handbook forSystematic Reviews of Interventions 4.2.6. Chichester, UK: JohnWiley & Sons, Ltd.; 2006.

17. The Cochrane Library. Section 6 Assessment of study quality.In: Higgins J, Green S, (eds). Cochrane Handbook for System-atic Reviews of Interventions 4.2.6. 4th eds. Chichester, UK:John Wiley & Sons, Ltd.; 2006.

18. Demark-Wahnefried W, Clipp EC, Lipkus IM, Lobach D,Snyder DC, Sloane R, et al. Main outcomes of the FRESHSTART trial: a sequentially tailored, diet and exercise mailedprint intervention among breast and prostate cancer survi-vors. J Clin Oncol. 2007;25:2709–2718.

19. De Bourdeaudhuij I, Stevens V, Vandelanotte C, Brug J.Evaluation of an interactive computer-tailored nutritionintervention in a real-life setting. Ann Behav Med. 2007;33:39–48.

20. Byrne N, Meerkin J, Laukkanen R, Ross R, Fogelholm M,Hills A. Weight loss strategies for obese adults: personalizedweight management program vs. standard care. Obesity.2006;14:177–188.

21. Campbell M, Demark-Wahnefried W, Symons M,Kalsbeek W, Dodds J, Cowan A, et al. Fruit and vegetableconsumption and prevention of cancer: the black churchesunited for better health project. Am J Pub Health. 1999;89:1390–1396.

22. Campbell MK, Tessaro I, DeVellis B, Benedict S, Kelsey K,Belton L, et al. Effects of a tailored health promotion programfor female blue-collar workers: health works for women.Prev Med. 2002;34:313–323.

23. Clutter Snyder D, Sloane R, Haines PS, Miller P, Clipp EC,Morey MC, et al. The Diet Quality Index-Revised: a tool topromote and evaluate dietary change among older cancersurvivors enrolled in a home-based intervention trial. J AmDiet Assoc. 2007;107:1519–1529.

24. Blalock SJ, DeVellis BM, Patterson CC, Campbell MK,Orenstein DR, Dooley MA. Effects of an osteoporosis preven-tion program incorporating tailored educational materials.Am J Health Promot. 2002;16:146–156.

25. Anderson ES, Winett RA, Wojcik JR, Winett SG, Bowden T.A computerized social cognitive intervention for nutritionbehavior: direct and mediated effects on fat, fiber, fruits, andvegetables, self-efficacy, and outcome expectations amongfood shoppers. Ann Behav Med. 2001;23:88–100.

26. Elder JP, Ayala GX, Campbell NR, Arredondo EM, Slymen DJ,Baquero B, et al. Long-term effects of a communication inter-vention for Spanish-dominant Latinas. Am J Prev Med.2006;31:159–166.

27. Heimendinger J, O’ Neill C, Marcus AC, Wolfe P, Julesburg K,Morra M, et al. Multiple tailored messages are effective inincreasing fruit and vegetable consumption among callers tothe Cancer Information Service. J Health Commun. 2005;10(Suppl1):65–82.

28. Jones H, Edwards L, Vallis TM, Ruggiero L, Rossi SR, Rossi JS,et al. Changes in diabetes self-care behaviors make a differ-ence in glycemic control: the diabetes stages of change(DiSC) study. Diabetes Care. 2003;26:732–737.

29. Kreuter MW, Sugg-Skinner C, Holt CL, Clark EM,Haire-Joshu D, Fu Q, et al. Cultural tailoring for mammogra-phy and fruit and vegetable intake among low-incomeAfrican-American women in urban public health centers.Prev Med. 2005;41:53–62.

Nutrition Reviews® Vol. 67(8):464–480478

30. Kristal AR, Curry SJ, Shattuck AL, Feng Z, Li S. A randomizedtrial of a tailored, self-help dietary intervention: the PugetSound Eating Patterns study. Prev Med. 2000;31:380–389.

31. Lutz SF, Ammerman AS, Atwood JR, Campbell MK,DeVellis RF, Rosamond WD. Innovative newsletter interven-tions improve fruit and vegetable consumption in healthyadults. J Am Diet Assoc. 1999;99:705–709.

32. Marcus AC, Heimendinger J, Wolfe P, Fairclough D, Rimer BK,Morra M, et al. A randomized trial of a brief intervention toincrease fruit and vegetable intake: a replication studyamong callers to the CIS. Prev Med. 2001;33:204–216.

33. Nitzke S, Kritsch K, Boeckner L, Greene G, Hoerr S, Horacek T,et al. A stage-tailored multi-modal intervention increasesfruit and vegetable intakes of low-income young adults. Am JHealth Promot. 2007;22:6–14.

34. Rothert K, Strecher VJ, Doyle L, Caplan WM, Joyce JS,Jimison HB, et al. Web-based weight management programsin an integrated health care setting: a randomized, controlledtrial. Obesity. 2006;14:266–272.

35. Sorensen G, Barbeau EM, Stoddard AM, Hunt MK, Goldman R,Smith A, et al. Tools for health: the efficacy of a tailored inter-vention targeted for construction laborers. Cancer CausesControl. 2007;18:51–59.

36. Tate DF, Jackvony EH, Wing RR. A randomized trial comparinghuman e-mail counseling, computer-automated tailoredcounseling, and no counseling in an internet weight lossprogram. Arch Intern Med. 2006;166:1620–1625.

37. Tilley BC, Glanz K, Kristal AR, Hirst K, Li S, Vernon SW, et al.Nutrition intervention for high-risk auto workers: results ofthe Next Step Trial. Prev Med. 1999;28:284–292.

38. Turnin MC, Beddok RH, Clottes JP, Martini PF, Abadie RG,Buisson JC, et al. Telematic expert system Diabeto. DiabetesCare. 1992;15:204–212.

39. Vandelanotte C, De Bourdeaudhuij I, Sallis J, Spittaels H,Brug J. Efficacy of sequential or simultaneous interactivecomputer-tailored interventions for increasing physical activ-ity and decreasing fat intake. Ann Behav Med. 2005;29:138–145.

40. Winett RA, Anderson ES, Bickley PG, Walbery-Rankin J,Moore JF, Leahy M, et al. Nutrition for a lifetime system: amultimedia system for altering food supermarket shoppers’purchases to meet nutritional guidelines. Comput HumBehav. 1997;13:371–392.

41. Winett RA, Anderson ES, Wojcik JR, Winett SG, Bowden T.Guide to health: nutrition and physical activity outcomes of agroup-randomized trial of an internet-based intervention inchurches. Ann Behav Med. 2007;33:251–261.

42. Wylie-Rosett J, Swencionis C, Ginsberg M, Cimino C,Wassertheil-Smoller S, Caban A, et al. Computerizedweight loss intervention optimizes staff time: the clinical andcost results of a controlled clinical trial conducted in amanaged care setting. J Am Diet Asso. 2001;101:1155–1162.

43. Anderson J, Dusenbury L. Worksite cholesterol and nutrition:an intervention project in Colorado. AAOHN J. 1999;47:99–106.

44. Prochaska JO, Velicer WF. The transtheoretical model ofhealth behavior change. Am J Health Promot. 1997;12:38–48.

45. Vandelanotte C, De Bourdeaudhuij I, Brug J. Two-yearfollow-up of sequential and simultaneous interactivecomputer-tailored interventions for increasing physical activ-ity and decreasing fat intake. Ann Behav Med. 2007;33:213–219.

46. Tate DF, Jackvony EH, Wing RR. Effects of internet behavioralcounseling on weight loss in adults at risk for type 2 diabetes:a randomized trial. JAMA. 2003;289:1833–1836.

47. Begg C, Berlin J. Publication bias: a problem in interpretingmedical data. J R Stat Assoc. 1988;151:419–463.

48. Begg C, Berlin J. Publication bias and dissemination ofclinical research. J Natl Cancer Inst. 1989;81:107–115.

49. Dickersin K, Min Y, Meinert C. Factors influencing pub-lication of research results: follow-up of applications sub-mitted to two institutional review boards. JAMA. 1992;263:374–378.

50. Schulz KF, Grimes DA. Multiplicity in randomised trials I: end-points and treatments. Lancet. 2005;365:1591–1595.

51. Hebert JR, Clemow L, Pbert L, Ockene IS, Ockene JK. Socialdesirability bias in dietary self-report may compromise thevalidity of dietary intake measures. Int J Epidemiol.1995;24:389–398.

52. Sterne JA, Egger M, Smith GD. Systematic reviews in healthcare: Investigating and dealing with publication and otherbiases in meta-analysis. BMJ. 2001;323:101–105.

53. Brug J, Campbell M, van Assema P. The application andimpact of computer-generated personalized nutrition educa-tion: a review of the literature. Patient Educ Couns.1999;36:145–156.

54. Ministry of Health. NZ Food: NZ People—Key Results of the1997 National Nutrition Survey. Wellington, New Zealand:Ministry of Health; 1999.

55. Ministry of Health. Food and Nutrition Guidelines for HealthyAdults: a Background Paper. Wellington, New Zealand: Minis-try of Health; 2003.

56. Australian Government and Ministry of Health. Nutrient Ref-erence Values for New Zealand and Australia: Including Recom-mended Dietary Intakes. Wellington, New Zealand: Ministry ofHealth; 2006.

57. U.S. Department of Health and Human Services, U.S. Depart-ment of Agriculture. Dietary Guidelines for Americans.Available at: http://www.health.gov/dietaryguidelines/dga2005/document/pdf/DGA2005.pdf, 2005. Accessed 27 March2009.

58. Casagrande SS, Wang Y, Anderson C, Gary TL. HaveAmericans increased their fruit and vegetable intake? Thetrends between 1988 and 2002. Am J Prev Med. 2007;32:257–263.

59. Wright JD, Wang CY, Kennedy-Stephenson J, Ervin RB.Dietary Intake of Ten Key Nutrients for Public Health, UnitedStates: 1999–2000. Advance data from vital and health sta-tistics; np.334. Hyattsville, Maryland: National Center forHealth Statistics; 2003.

60. Moher D, Schulz KF, Altman DG, Group C. The CONSORTstatement: revised recommendations for improving thequality of reports of parallel-group randomized trials. J AmPodiatr Med Assoc. 2001;91:437–442.

61. Davidson KW, Goldstein M, Kaplan RM, Kaufmann PG,Knatterud GL, Orleans CT, et al. Evidence-based behavioralmedicine: what is it and how do we achieve it? Ann BehavMed. 2003;26:161–171.

62. Block Dietary Data Systems. Block 95 Food FrequencyQuestionnaire. Berkley, CA: Block Dietary Data Systems;1995.

63. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J,Gardner L. A data-based approach to diet questionnairedesign and testing. Am J Epidemiol. 1986;124:453–469.

64. Thompson FE, Byers T. Dietary assessment resource manual.J Nutr. 1994;124(Suppl 11):S2245–S2317.

Nutrition Reviews® Vol. 67(8):464–480 479

65. Vandelanotte C, Matthys C, De Bourdeaudhuij I. Reliabilityand validity of a computerised questionnaire to measure fatintake in Belgium. Nutr Res. 2004;24:621–631.

66. Subar AF, Thompson FE, Kipnis V, et al. Comparative valida-tion of the Block, Willett, and National Cancer Institute foodfrequency questionnaires: the Eating at America’s TableStudy. Am J Epidemiol. 2001;154:1089–1099.

67. Block G, Woods M, Potosky A, Clifford C. Validation of a self-administered diet history questionnaire using multiple dietrecords. J Clin Epidemiol. 1990;43:1327–1335.

68. Berkeley Nutrition Services. Block 98 Food FrequencyQuestionnaire. Berkeley, CA.: Berkeley Nutrition Services;1998.

SUPPORTING INFORMATION

Additional Supporting Information may be found in theonline version of this article:

Table S1: Characteristics and results of all 25 trialsmeeting inclusion criteria for the review.

Please note: Wiley-Blackwell are not responsible for thecontent or functionality of any supporting materials sup-plied by the authors. Any queries (other than missingmaterial) should be directed to the corresponding authorfor the article.

Nutrition Reviews® Vol. 67(8):464–480480