McGill, R; Anwar, E; Orton, L; Bromley, H; Lloyd-Williams, F

16 Pages·2017·1.69 MB·English

McGilletal.BMCPublicHealth (2015) 15:457
DOI10.1186/s12889-015-1781-7
RESEARCH ARTICLE Open Access
Are interventions to promote healthy eating
equally effective for all? Systematic review of
socioeconomic inequalities in impact
RoryMcGill1*,ElspethAnwar1,LoisOrton1,HelenBromley1,Ffion Lloyd-Williams1,MartinO’Flaherty1,
DavidTaylor-Robinson1,MariaGuzman-Castillo1,DuncanGillespie1,PatriciaMoreira1,KirkAllen1,LirijeHyseni1,
NicolaCalder1,Mark Petticrew2,MartinWhite3,4,MargaretWhitehead1and SimonCapewell1
Abstract
Background: Interventionstopromotehealthyeatingmakeapotentiallypowerfulcontributiontotheprimaryprevention
ofnoncommunicablediseases.Itisnotknownwhetherhealthyeatinginterventionsareequallyeffectiveamongall
sectionsofthepopulation,norwhethertheynarroworwidenthehealthgapbetweenrichandpoor.
Weundertookasystematicreviewofinterventionstopromotehealthyeatingtoidentifywhetherimpactsdifferby
socioeconomicposition(SEP).
Methods:Wesearchedfivebibliographicdatabasesusingapre-pilotedsearchstrategy.Retrievedarticleswerescreened
independentlybytworeviewers.Healthierdietsweredefinedasthereducedintakeofsalt,sugar,trans-fats,saturatedfat,
totalfat,ortotalcalories,orincreasedconsumptionoffruit,vegetablesandwholegrain.Studieswereonlyincludedif
quantitativeresultswerepresentedbyameasureofSEP.
Extracteddatawerecategorisedwithamodifiedversionofthe“4Ps”marketingmix,expandedto6“Ps”:“Price,Place,
Product,Prescriptive,Promotion,andPerson”.
Results:Oursearchidentified31,887articles.Followingscreening,36studieswereincluded:18“Price”interventions,6
“Place”interventions,1“Product”intervention,zero“Prescriptive”interventions,4“Promotion”interventions,and18
“Person”interventions.
“Price”interventionsweremosteffectiveingroupswithlowerSEP,andmaythereforeappearlikelytoreduceinequalities.
Allinterventionsthatcombinedtaxesandsubsidiesconsistentlydecreasedinequalities.Conversely,interventionscategorised
as“Person”hadagreaterimpactwithincreasingSEP,andmaythereforeappearlikelytoreduceinequalities.Allfour
dietarycounsellinginterventionsappearlikelytowideninequalities.
Wedidnotfindany“Prescriptive”interventionsandonlyone“Product”interventionthatpresenteddifferentialresults
andhadnoimpactbySEP.More“Place”interventionswereidentifiedandnoneoftheseinterventionswerejudgedas
likelytowideninequalities.
Conclusions:Interventionscategorisedbya“6Ps”frameworkshowdifferentialeffectsonhealthyeatingoutcomesby
SEP.“Upstream”interventionscategorisedas“Price”appearedtodecreaseinequalities,and“downstream”“Person”
interventions,especiallydietarycounsellingseemedtoincreaseinequalities.
HoweverthevastmajorityofstudiesidentifieddidnotexploredifferentialeffectsbySEP.Interventionsaimedatimproving
populationhealthshouldberoutinelyevaluatedfordifferentialsocioeconomicimpact.
Keywords:Noncommunicablediseases,Socioeconomicinequalities,Healthyeating,Intervention
*Correspondence:[email protected]
1DepartmentofPublicHealthandPolicy,UniversityofLiverpool,Liverpool,
UK
Fulllistofauthorinformationisavailableattheendofthearticle
©2015McGilletal.;licenseeBioMedCentral.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative
CommonsAttributionLicense(http://creativecommons.org/licenses/by/4.0),whichpermitsunrestricteduse,distribution,and
reproductioninanymedium,providedtheoriginalworkisproperlycredited.TheCreativeCommonsPublicDomain
Dedicationwaiver(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthisarticle,
unlessotherwisestated.
McGilletal.BMCPublicHealth (2015) 15:457 Page2of15
Background available which target the components outlined above -
Non communicable diseases (NCD’s e.g. cardiovascular cost,accessandknowledge.Furthermore,suchpopulation
disease (CVD), chronic obstructive pulmonary disease, interventions, by their very nature, should theoretically
diabetes, cancer, etc.) remain the major cause of disease, benefiteveryoneinthepopulation,includingthosewitha
disability and death, accounting for over 63% of deaths historyofNCDsuchasCVD.
worldwide in 2012 [1]. A substantial amount of the NCD However, there is a lack of evidence concerning the
burdenisattributabletofourbehaviouralriskfactors(not- health equity impact of dietary interventions to promote
ablypoordiet,alsosmoking, alcohol and physicalinactiv- health. This has led to an increase in systematic reviews
ity). Poor nutrition causes a greater population burden of assessing health equity effects [14,15]. Preventive inter-
morbidityandmortalityfromNCDsthantobacco,alcohol ventions may not benefit all sub groups of the popula-
and physical activity combined [2]. Furthermore, the tion equally [16,17]. This has been termed “intervention
prevalence of NCD risk factors and hence burden of generated inequalities”or“IGIs”[18].
NCDs are not equally distributed throughout the popula- White et al. have described the points in the imple-
tion [3]. There is evidence for an inverse relationship be- mentation of an intervention which may impact upon
tween socioeconomic position (SEP) and most risk differential effectiveness by SEP [18]. These include
factors, withNCD riskfactors oftenbeinghigherin more intervention efficacy, service provision or access, uptake,
disadvantagedgroups(lowSEP)[3]. and compliance [15]. Compliance may be higher among
Thus, eating a healthy diet demonstrates a social gra- more advantaged groups because of better access to re-
dient with diet among people in lower SEPs being sources such as time, finance, and coping skills. “Down-
poorer in quality when compared to more advantaged stream” interventions (which rely solely on individuals
groups. The World Health Organisation (WHO) define making and sustaining behaviour change) may therefore
a healthy diet as achieving energy balance, limiting en- be more likely to be taken up by those who are of higher
ergy intake from total fats, free sugars and salt and in- SEP and are more likely to widen the health gap between
creasing consumption of fruits and vegetables, legumes, rich and poor. Conversely, those of lower SEP tend to be
wholegrainsandnuts[4]LowerSEPisassociated witha harder to reach, and find it harder to change behaviour
higher intake of energy dense, nutrient poor foods duetoalackofaccesstotheresourcespreviouslyoutlined
(which are high in saturated fat and sugar), and with [19].“Upstream”interventionsremovethisrelianceonre-
lowerintakeoffruit,vegetablesandwholegrains [5]. source availability. Due to a higher risk burden, those of
Socioeconomic inequalities in diet are influenced by lower SEPare likely togainextrabenefit if a riskfactor is
factorsincludingcost,accessandknowledge.Adietrela- uniformlyreducedacrosstheentirepopulation.Therefore
tively high in energy is generally less expensive than a beingmorelikelytoreduceinequalities[16,20].
diet consisting of less energy dense products, such as Thomas and colleagues demonstrated differential im-
vegetables [6]. Food selection is not only a behavioural pact of tobacco control policy interventions. They
choice, but also an economic one [7]. Access to healthy showed that population level tobacco control interven-
foods can also be inequitable. This can be a lack of tions, such as increasing the price of tobacco products
healthy food options provided in shops within disadvan- had a greater potential to benefit more disadvantaged
taged areas [8] which has been described in the US in groups and thereby reduce health inequalities [17]. With
terms of “food deserts”, however evidence for these have deprived groups already having a higher NCD burden
not been found within other settings e.g. UK [9]. Signifi- (in 2008 worldwide age standardised mortality rates
cant differences in nutritional knowledge have been from NCDs were almost twice as high for lower income
shown between differing socioeconomic groups, with groups when compared to higher income groups [1]),
knowledge declining with lower socioeconomic status there is an urgent need to further explore this important
[10]. In children, lower SEP is associated with a subse- issue relatingtothemajorNCDrisk factor,diet[2,21].
quentincreasedriskofadultcardiovascularmorbidityand Oldroyd and colleagues [22] previously examined the
mortality,partlyreflectinglowerexposuretohealthyfoods differential effects of healthy eating interventions by
[11]. This can then reinforce adult food preferences for relative social disadvantage. In their small number of in-
lesshealthyfoods[12]. cluded studies they found limited evidence of greater
There has been considerable effort to develop impact in less disadvantaged groups [22]. This may be
population-widedietaryinterventions.Theseprimarypre- due to their chosen time frame (1990–2007) and limited
vention programmes are aimed at asymptomatic individ- databasessearched(MEDLINE andCINAHL).
uals in the normal population, before any negative health Our aim was to update and expand upon Oldroyd and
eventhasoccurred[13].Interventionsatthisstageaimto colleagues review [22]. In order to identify interventions
modify NCD risk factors through the promotion of which may reduce inequalities in healthy eating, we
healthier diets. Potentially powerful interventions are undertook a systematic review of interventions (and
McGilletal.BMCPublicHealth (2015) 15:457 Page3of15
modelling studies) to promote healthy eating in general systematic reviews that were identified) were scrutinised
populations,todeterminewhetherimpactsdifferbySEP. forotherpotentiallyeligible studies.
Methods
Studyselectionandinclusioncriteria
Studydesign
We included studies of any design that assessed the ef-
We conducted a systematic review with a combination of
fectsofinterventionstopromotehealthyeating(reduced
graphical and narrative synthesis of published literature.
intake of salt, sugar, trans-fats, saturated fat, total fat, or
We followed best practice guidance as detailed by the
total calories, or increased consumption of fruit, vegeta-
PRISMA-Equity 2012 Extension for systematic reviews
bles andwholegrain)targeted athealthypopulations that
withafocusonhealthequity.Thistoolhasbeendescribed
reported quantitative outcomes by a measure of SEP.
as a method to improve both the reporting and conduct
Only studies published since 1980 in the English lan-
of equity focused systematic reviews [23] (provided in the
guage were considered. Upon fulfilling these criteria,
additionalinformation–Additionalfile1).
studies were assessed utilising a PICOS (Participants, In-
terventions, Comparators, Outcomes and Study design)
Searchstrategy [23].This issummarisedinTable1.
In order to identify all relevant studies, a pre-piloted One reviewer (RMcG) screened titles, removed dupli-
search strategy was used to search five bibliographic da- cates and selected potentially relevant abstracts. Then
tabases (MEDLINE, Psycinfo, SCI, SSCI and SCOPUS). two reviewers(RMcG&EA) independentlyexaminedall
An example of the search strategy used is provided in the abstracts for eligibility. All articles deemed poten-
the additional information (Additional file 2). In tially eligible were retrieved in full text. The full text was
addition, we screened titles from the reference sections also retrieved for any abstracts where a decision could
of systematic reviews in the Campbell library, CEN- not be made based on the information given. Full text
TRAL, DARE and EPPI. Colleagues and experts from articles were then screened independently by the two re-
key organisations working in public health policy were viewers (RMcG&EA).Disagreementsoneligibility deci-
alsocontacted foranyadditionaldata sources. Therefer- sions were resolved by consensus or by recourse to a
ence lists of all included studies (including relevant senior memberofthereviewteam (SC).
Table1PICOSapproachtostudyeligibility*
Include Exclude
Participants
Healthypopulations(anyageorgender),fromanycountry Studiesincludingparticipantsthatwerenotrepresentativeofthe
populationwereexcluded(e.g.subcategoriessuchasobeseparticipantsin
weightlosstrials,participantswithdiabetes,pregnantwomen).
Interventions
Studiesevaluatingtheeffectsofinterventiontopromotehealthyeating Interventionswithnochangeinhealthyeatingoutcomesquantitatively
thatwereimplementedexperimentally;orduetolocalornational stratifiedbySEP.
policies.Thesecouldincludearangeofactionstoimprovehealthy
Actionsinitiatedbyindustry.
eating(intermsofthedietaryfactorsofsalt,sugar,transfats,saturated
fat,totalfat,fruitandvegetablesandcalories).
Comparators
Studieswereonlyincludedinthereviewprovidedthattheauthorsmadea Studieswhichdidnotreporttheeffectsofactionstoimprovehealthy
quantitativecomparisonofdifferentialeffectsofpolicyinterventionsto eatingbySEP
improvehealthyeatingbyatleastonemeasureofSEP.
Outcomes
Theprimaryoutcomeofinterestwasdietaryintake.Secondaryoutcomes Processevaluationsreportingonimplementationofinterventions/policies
included:changesinclinical/physiologicalindicatorsrelatedtoNCD,behaviours withoutanyoutcomedata;dataonlyoncosts,orfeasibilityoracceptability
associatedwithahealthydiete.g.changeinBMI. withoutanassessmentofintake;reviews/studiesofunder-nutrition.
StudieswithnomentionofSEP.
Studydesign
Weincludedstudiesofanydesign,includingRCTs,cohortstudiesand Opinionarticles;purelyqualitativeevaluationswithnoquantitative
modellingstudies.Weexplicitlyincludedmodellingstudiestobetter assessment;data/statisticsfrommonitoringandsurveillancenotdirectly
captureanalysisoffiscalmeasuressuchastaxes,subsidies,oreconomic linkedtoapolicyintervention
incentives
*PICOS=Participants,Interventions,Comparators,OutcomesandStudydesign.
McGilletal.BMCPublicHealth (2015) 15:457 Page4of15
Dataextractionandmanagement The six intervention categories used in the analysis are
Data from all included studies were extracted by one re- thus:
viewer using pre-designed and piloted forms. The ex-
tracted data was then checked independently by a (cid:1) Price–fiscalmeasuressuchastaxes,subsidies,or
second reviewer to ensure all the correct information economicincentives
was recorded. Extracted data included: study design, (cid:1) Place –environmentalmeasures inspecific settings
aims, methodological quality, setting, participants, and suchasschools,workplaces(e.g. vending machines)
outcomes related to thereview objectives.Extracted data orplanning(e.g.locationofsupermarketsandfast
were compared for accuracy and completeness. Where foodoutlets)orcommunity-basedhealtheducation
more information was required from an identified art- (cid:1) Product–modificationoffoodproductstomake
icle,theauthors were contacted where possible. them healthier/less harmfule.g.reformulation,
The measurement of SEP within the intervention was additives,oreliminationofa specificnutrient
carefully noted and included: education level, level of (cid:1) Prescriptive –restrictions onadvertising/marketing
household income, occupational status and ethnicity, as through controls orbans,labelling,
determinedbytheauthors[24,25].Ethnicitywasonlyin- recommendationsorguidelines
cluded as a measure of SEP if the authors explicitly (cid:1) Promotion–massmediapublic information
stated this was their SEP measurement proxy within the campaigns
text. Ifnot,weassumedthattheseweremeasuresofcul- (cid:1) Person–Individual-basedinformation and education
tural differences rather than socioeconomic inequalities (e.g.cookinglessons,tailorednutritionaleducation/
and these were excluded from the main analysis [26].In- counselling,ornutritioneducationintheschool
terventions targeting only deprived groups were not in- curriculum).
cluded as these did not include a comparison of the
effects of an intervention with higher SEP. All data ex- Socioeconomicinequalitiesinimpact
traction tables are included inthe additional information For each of the included interventions, if the outcome
(Additional file3). was split by more than one socioeconomic proxy meas-
ure, we took the quantitative effect on inequalities from
Assessmentofmethodologicalqualityofincludedstudies thestratifiedresultsthatbestrepresentedSEP[24,25].
The methodological quality of each included study was When calculating the effect on inequalities, we exam-
assessedindependentlybytworeviewersusingthecriteria ined the primary outcome of interest for each interven-
fortheCommunityGuideoftheUSTask ForceonCom- tion as identified by the study author. If a change in
munity Preventive Services and a six-item checklist of dietary intake was given this was the primary measure
quality of execution adapted from the criteria developed that was used. If not, some other secondary outcomes
for the Effective Public Health Practice Project [27,28]. were acceptable (see Table 1). We compared the lowest
Several of the included studies were modelling studies. groupwiththehighestgroupintheSEPclassification,and
Since these studies could not be assessed using the same used the measuresofsignificance reported bythe authors
qualityassessmenttoolastheempiricalstudies,twomod- (e.g. p values, confidence intervals, standard deviations,
elling experts assessed the quality of these independently. standard error of measurement) to assess the significance
Disagreements in methodological quality assessment for of any differential effects of interventions by SEP. When
all the included studies were resolved by consensus or by the results were stratified by age, gender or intervention
recoursetoaseniormemberofthereviewteam. site, the results referring to the largest subsample were
used. Where information was given at different time
Datasynthesis points,thelongestfollowupperiodwasexamined.
Weexaminedtheevidenceaboutthedifferentialeffectsof Theeffectoninequalitieswasclassifiedasfollows:
interventions in terms of their underlying theories of
change [29]. Different frameworks have been proposed to (cid:1) Intervention likelytoreduce inequalities:the
categorise healthy eating interventions [30]. However no interventionpreferentially improvedhealthyeating
one framework has been used consistently. The “4 Ps” outcomes inpeopleoflowerSEP
framework is a well-established framework used within (cid:1) Intervention likelytowiden inequalities:the
the marketing field and translates well to a policy context interventionpreferentially improvedhealthyeating
[31]. This framework includes interventions examining outcomes inpeopleof higher SEP
“Price”, “Place”, “Product” and “Promotion”. We have (cid:1) Intervention which had nopreferentialimpactby
adapted and strengthened this framework in orderto cat- SEP(thisalsoincludesinterventions wherethere
egorise policy interventions relating to healthy eating by wasanoverallbenefitbutwhere there wasnoeffect
theirmechanismsofunderlyingchange. onhealthyeatingoutcomesforanySEPsub-group).
McGilletal.BMCPublicHealth (2015) 15:457 Page5of15
We aspiredto undertakeameta-analysis of theresults. preferential impact by SEP), or were less effective in dis-
However the studies identified were heterogeneous, ad- advantagedgroups(widen)(Figure 1).
dressing different research questions, with diverse theor-
etical underpinnings study designs and study outcomes.
Given the considerable heterogeneity of the studies, Sensitivityanalyses
undertaking a meta-analysis was not deemed appropri- We conducted a sensitivity analysis to determine if the
ate. The results were therefore synthesised using a com- key resultswould changeifwehadbeenmore orlessse-
bination of graphical and narrative methods, including lective inourstudyscreeningprocess.
the use of the Harvest plot, which is a useful graphical First, we included only the studies which gave indica-
method for synthesising and displaying evidence about tors ofstatisticalsignificance concerning the quantitative
the differential effects of population-level interventions data split by SEP. Secondly, we also included those stud-
[32]. Within the Harvest Plot, each intervention was ies which split their findings quantitatively by ethnicity
represented as a single bar in one of three categories: alone (with no mention of SEP), as this represents a
those that were more effective in more disadvantaged crude proxy measure of SEP [33] (see additional infor-
groups (reduce), had the same effect in all groups (no mation -Additionalfile4).
Figure1HarvestPlotsummarisingtheeffectsofhealthyeatinginterventionsoninequalities*.*EachmatrixwithintheHarvestplot‘supermatrix’
illustratesourfindingsforeach“P”.Eachmatrixconsistsofthreecolumnsindicatingwhetherinequalitieswerereduced,widenedorshowedno
gradient.Eachbarrepresentsoneintervention.Theheightofthebarindicatesthequalityscoreofthestudygradedoutof6[28].Greybarsindicate
interventionswithnosignificancevaluesgivenconcerningthedifferenceineffectoftheinterventiononSEP.Modellingstudiesareindicated
bypatternedbars
McGilletal.BMCPublicHealth (2015) 15:457 Page6of15
Results the bar depicts the quality of the study. Modelling stud-
We identified 31,887 articles in our search. Following ab- iesweredistinguishedbyusingpatterned bars.
stractandfulltextscreening,36studiesmettheinclusion The studies are then grouped by outcome regarding
criteria(Figure2).Theseincludedquantitativeresultspre- socioeconomic differential effects (reduced, no preferen-
sented by a measure of SEP for 47 interventions. A sum- tial impact by SEP and widened). Interventions in the
mary of all included studies is listed in the additional “Price” category appeared most likely to reduce inequal-
information(Additionalfile5).Dataextractiontablesfor ities while “Person” interventions were the most likely to
all included studies and studies included in the sensitiv- widen inequalities(Figure 1).
ity analysis are provided in the additional information
(Additionalfile3). Priceinterventions(taxes,subsidies,oreconomicincentives)
Eighteen “Price” interventions were identified. These
Impactonsocioeconomicinequalitiesby“P”category are summarised in Table 2. The majority were con-
The impact of interventions categorised by “P” is dis- ducted in Europe [34-39], with five in North America
played in the Harvest plot in Figure 1 (adapted from [40,41] and one in Australia [42]. Of these, nine were
Thomas et al. [17]). The Harvest plot shows each inter- taxes on high energy density foods [34,36,37,41,42],
vention illustrated as an individual bar. The height of three were subsidies on fruit and vegetables [35,40]
Figure2Flowchartshowingtheprogressofthereview.*studiessumto37becauseonestudyexaminedseveraldifferenttypesofinterventions
whichwereincludedintwoseparatecategories.
McGilletal.BMCPublicHealth (2015) 15:457 Page7of15
Table2Summaryof“Price”interventions
Author Study Setting Intervention QualityΔ Outcome SEP EffectonSEP
measured measurement inequalities†
Allais[34] Modelling France 10%Taxonhighenergydensityfood: 2 Changeinfat Household ↓
study consumed(%) income
Cash[40] Modelling USA 1%Subsidyonfruitandvegetables 2 CHDincidence Household ↑*
study income
Dallongeville[35] Modelling France 5.5%to2.1%Subsidyonfruitand 2 Changeinmean Household ↔*
study vegetables fruitandvegetable income
consumption(g/d)
Foodstampprogramforfruitand ↓*
vegetables
Finkelstein[41] Modelling Canada 20%Taxonhighenergydensityfood 2 Meanchangein Household ↔*
study energyintakefrom income
40%taxoncarbonatedsugar ↔*
allbeverages
sweetenedbeverages
20%taxonallsugarsweetened ↔*
beverages
40%taxonallsugarsweetened ↔*
beverages
Nederkoorn[36] RCT Holland 50%Taxonhighenergydensityfood 5 %changein Foodbudget ↓*
caloriespurchased
inleanindividuals
Nnoaham[37] Modelling UK 17.5%taxonhighenergydensityfoods 2 %changeincalorie Household ↔*
study intake income
17.5%taxonfoodclassifiedas‘less ↓*
healthy’bynutrientprofiling
Combinedthetaxationon‘lesshealthy’ ↓*
foodswitha17.5%subsidyonfruitand
vegetables
Asabovewitha32.5%subsidyonfruit ↓*
andvegetables
Sharma[42] Modelling Australia 20%taxonsugarsweetenedbeverages 2 Meannetchangein Household ↑*
study bodyweightinkg income
Smed[38] Modelling Denmark 5%taxonfattymeatanddairy 2 Changeinnutrient Socialclass ↓
study productswithsubsidiesonfruitand demandof
vegetables,potatoesandgrain saturatedfat(%)
products
7.89DKK/kgtaxonsaturatedfatswith ↓
subsidiesonfibre
7.89DKK/kgtaxonsaturatedfatswith ↓
subsidiesonfibrewithanadditional
10.3DKK/kgtaxonsugar
Tiffin[39] Modelling UK 1%Taxonfattyfoodforevery% 2 %changeinenergy Occupation ↓
study saturatedfatcontentwithamatching intake
subsidyonfruitandvegetables
ΔQualityofempiricalstudieswereassessedusingavalidatedtool[27].Studieswerescoredagainstsixcriteriaandthisnumberwassummedtogiveanoverall
qualityscore(maximumofsix).Themodellingstudieswereassessedforqualitybytwoindependentexpertsandtheirscoreswereconvertedintoascoreoutof
sixtoallowcomparison.
†theeffectoninequalitiesisdisplayedsymbolicallyinthetableas:↓foranInterventionlikelytoreduceinequalities:theinterventionpreferentiallyimproved
healthyeatingoutcomesinpeopleoflowerSEP,↑foraninterventionlikelytowideninequalities:theinterventionpreferentiallyimprovedhealthyeating
outcomesinpeopleofhigherSEP,and↔foraninterventionwhichhadnopreferentialimpactbySEP.
*indicatesinterventionswherestatisticalsignificancevaluesweregiventothequantitativeevidencerelevanttoourreview.
and six were combinations of taxes and subsidies combination of taxes and subsidies consistently had a
[37-39]. Eight studies used modelling methodologies greater impact on lower SEP [37-39]. Two interventions
[34,35,37-42]. (one subsidy on fruit and vegetables [40] and one tax on
In total, ten of the eighteen “Price” interventions were high energy density foods [42]) had a greater impact on
likely to reduce inequalities by preferentially improving higher SEP, and there was no differential effect demon-
healthy eating outcomes in lower SEPs [34-39]. All six strated in the remaining six studies in the “Price” cat-
studies reporting interventions which consisted of a egory [35,37,41].
McGilletal.BMCPublicHealth (2015) 15:457 Page8of15
Placeinterventions(environmentalmeasuresinspecific [44,45], one schools based intervention [46] and one
settings) areabasedintervention [47]).
Six “Place” interventions were identified. These are sum-
marisedinTable3.ThreewerecarriedoutinNorthAmerica
[43-45],twoinEurope[46,47]andoneinNewZealand[48]. Productinterventions(modificationoffoodproductsto
Of these, two were school based interventions [46,48], two makethemhealthier/lessharmful)
were work based interventions [44,45], one church based Only one “Product” intervention was identified [49].
intervention[43]andoneareabasedintervention[47]. This intervention is summarised in Table 3. This was a
None of the six identified “Place” interventions were product reformulation intervention conducted inthe UK
judged as likely to widen inequalities, with four likely to (salt) inwhich the authors identified no impactbysocio-
reduce inequalities (both work place interventions economicgradient.
Table3Summaryof“Place”,“Product”,“Prescriptive”and“Promotion”interventions
Author Study Setting Intervention QualityΔ Outcomemeasured SEP EffectonSEP
measurement inequalities†
Place
Campbell[43] RCT USA Churchbased 5 Meanchangeinportionsof Household ↔*
intervention fruitandvegetablesconsumed income
Hughes[46] Cross England Schoolbased 4 Changeinportionsoffruitand Indexof ↓*
sectional intervention vegetablesconsumed Multiple
survey Deprivation
Rush[48] RCT New Schoolbased 3 ChangeinBMIstandard Household ↔*
Zealand intervention deviationscorein5–7year income
olds
Sorenson[44] RCT USA Workbasedintervention 5 %changeinthoseachieving5 Occupation ↓*
aday
Sorenson[45] RCT USA Workbasedintervention 5 Changeingeometricmean Occupation ↓*
gramsoffibreper1000kcals
Wendel-Vos[47] Cohortstudy Holland Areabasedintervention 4 Differenceinmeanenergy Education ↓*
intakebetweenintervention level
andcontrol(MJ/d)
Product
Millet[49] Observational England Saltreformulation 3 Saltintake(g/d) Socialclass ↔*
study
Prescriptive
NostudieswereidentifiedexaminingthepotentialSEPdifferentialseffectsofrestrictionsonadvertising/marketingthroughcontrolsorbans;labelling,
recommendationsorguidelines
Promotion
Cappacci[50] Modelling UK Healthinformation 2 Changeinfruitandvegetable Household ↓*
study campaign(5aday) intake(portions) income
Dallongeville[35] Modelling France Healthinformation 2 Changeinfruitandvegetable Household ↔*
study campaign(fruitand consumption(g/d) income
vegetablepromotion)
Estaquio[51] Cohortstudy France Healthinformation 2 %ofmalesconsuming≥five Education ↑*
campaign(5aday) portionsoffruitandvegetable level
perday
Stables[52] Cross USA Healthinformation 2 Changeinportionsoffruitand PovertyIndex ↔*
sectional campaign(5aday) vegetablesconsumed Ratio
survey
ΔQualityofempiricalstudieswereassessedusingavalidatedtool[27].Studieswerescoredagainstsixcriteriaandthisnumberwassummedtogiveanoverall
qualityscore(maximumofsix).Themodellingstudieswereassessedforqualitybytwoindependentexpertsandtheirscoreswereconvertedintoascoreoutof
sixtoallowcomparison.
†theeffectoninequalitiesisdisplayedsymbolicallyinthetableas:↓foranInterventionlikelytoreduceinequalities:theinterventionpreferentiallyimproved
healthyeatingoutcomesinpeopleoflowerSEP,↑foraninterventionlikelytowideninequalities:theinterventionpreferentiallyimprovedhealthyeating
outcomesinpeopleofhigherSEP,and↔foraninterventionwhichhadnopreferentialimpactbySEP.
*indicatesinterventionswherestatisticalsignificancevaluesweregiventothequantitativeevidencerelevanttoourreview.
McGilletal.BMCPublicHealth (2015) 15:457 Page9of15
Prescriptiveinterventions(restrictionsonadvertising/ position (SEP). “Upstream” interventions categorised as
marketing) “Price” appeared most likely to decrease health inequal-
No“Prescriptive”interventionswereidentified. ities,while“downstream”“Person”interventionsappeared
most likely to increase inequalities (this association weak-
Promotioninterventions(massmediapublicinformation enedwhenonlystudieswhichreportedsignificancevalues
campaigns) pertainingtoSEPdifferentialeffectivenesswereincluded).
Four “Promotion” interventions were identified. These No “Prescriptive” interventions were found and only one
are summarised in Table 3. Three of these were con- intervention categorised as “Product” was included.
ducted in Europe [35,50,51] and one in the USA [52]. “Place” interventions showed mixed results, although
All four examined the effectiveness of national “Five a none appeared likely to widen inequalities. However, the
day” health information campaigns. Two studies used vast majority of full text articles which were assessed for
modellingmethodologies [35,50]. eligibilitydidnotexploredifferentialeffectsbySEP.
“Promotion” interventions showed mixed results. Two
interventions had no preferential impact by SEP [35,52] Comparisonwithotherresearch
while one intervention was judged as likely to reduce in- This research builds on an earlier systematic review by
equalities [50] and the other intervention judged as Oldroyd and colleagues who examined effectiveness of
likely towiden inequalities[51]. nutrition interventions on dietary outcomes by relative
social disadvantage [22]. They concluded that nutrition
Personinterventions(Individual-basedinformationand interventions have differential effects, but could not de-
education) velop this further due to the small number of studies
Eighteen“Person”interventionswereidentified.Theseare identified. Our review included 36 studies allowing ex-
summarised in Table 4. The majority of these were con- pansion upon these conclusions. Magnée et al. has re-
ducted in Europe [53-61], eight in the USA [62-68] and cently used a systematic approach exploring the
one in Australia [69]. Of these, fourteen were health edu- socioeconomic differential impact of lifestyle interven-
cationinterventions[53-56,58-60,62,63,65,67-69]andfour tions (including diet) related to obesity prevention in a
weredietarycounsellinginterventions[57,61,64,66]. Dutchsetting [78].They tooreported that“downstream”
“Person” interventions were judged as most likely to interventions targeting individuals might increase in-
widen inequalities, with eight of the eighteen interven- equalities but their findings were limited by a lack of
tions having greater impact in higher SEPs studiesexaminingsocioeconomicdifferential effects.
[57,59-61,64-66,68]. All four of the dietary counselling Why might “Price” and “Person” interventions affect
interventionsappearlikely towiden inequalities. inequalities differently? White et al. suggest that how an
interventionisdeliverediscrucial.Hencestructural,uni-
Sensitivityanalysis versally delivered “upstream” interventions which create
When the screening process was made more selective, a healthier environment therefore tend to circumvent
the general trends seen in the main Harvest plot were voluntary behaviour change may well reduce inequalities
essentially unchanged. “Price” interventions remained [18]. Frieden depicts this difference as a “Health Impact
the most likely to reduce inequalities, however “Person” Pyramid” [79].Thebase of the pyramid consists ofinter-
interventions now showed mixed results with a more ventions addressing socio-economic determinants of
even distribution of effects by SEP when being more se- health which has the greatest potential population im-
lective by only including interventions where statistical pact. Conversely, the top of the pyramid depicts health
significance valuesweregiven.Therewerenodifferences education and counselling which depend on higher
observed related to the other “P” categories. The levels of individual effort; hence resulting in the lowest
addition of studies that split their findings by ethnicity potential population impact. Cappuccio and colleagues
alone [70-77] (making the selection process less select- likewise found that more “upstream” population-wide
ive) had no implications on the main findings (see add- regulation and marketing controls had the most poten-
itional information – Additional file 4). Six of these tial to reduce dietary salt when compared with more
studies were from the USA, with one from New Zealand “downstream”approacheslike foodlabelling [80].
andonefrom theNetherlands. Our review supports both White and Frieden [18,79].
Interventions in the “Price” category predominantly in-
Discussion cluded taxes on unhealthy foods and subsidies for
Mainfindings healthier foods; both are population level, structural in-
Interventionscategorisedbythe“6Ps”modifiedversionof terventionswhichrequirenoindividualagency.Thiscat-
the “marketing mix” framework demonstrated differential egory was the most likely to reduce inequalities. Similar
effects on healthy eating outcomes by socioeconomic observations have also been demonstrated for tobacco
McGilletal.BMCPublicHealth (2015) 15:457 Page10of15
Table4Summaryof“Person”interventions
Author Study Setting Intervention QualityΔ Outcomemeasured SEP EffectonSEP
measurement inequalities†
Brownson[62] Cross USA Healtheducation: 3 %changeofthe%ofpeople Education ↓
sectional Communitybased whoconsumefiveportionsof level
survey education fruitandvegetablesperday
Burgi[53] RCT Switzerland Healtheducation:Healthy 5 MeanBMI(kg/m2) Parental ↔*
nutritionprogramaimedat education
children level
Carcaise- RCT USA Healtheducation:Tailored 5 Meanfruitandvegintake Education ↓*
Edinboro[63] feedbackandself-helpdiet- score(Scoreoutof3,3=less level
aryintervention. F/Vintake,1=moreF/Vintake)
Connett[64] RCT USA Dietarycounselling 3 Changeinserumcholesterol Household ↑
intervention (mg/dl) income
Curtis[54] Randomised UK Healtheducation:Cooking 3 %changeinmeanfood Quintileof ↓*
parallel fairwithcookinglessons energyfromfat Deprivation
groups accompanying Index
comparison personaliseddietarygoal
study settings
Friel[55] RCT Republicof Healtheducation:Healthy 2 Changein%ofchildren Arealevel ↔*
Ireland nutritionprogramaimedat consuming>4portionsoffruit deprivation
children(“Heartyheart”) andvegperday
Haerens[56] RCT Belgium Healtheducation:adapted 4 Changeinmeandietaryfat Education ↔*
computertailoreddietary intake(g/d) level
interventionforchildren.
Havas[65] RCT USA Healtheducation:Healthy 5 Changeinmeandailyservings Education ↑*
nutritionprogramaimedat consumedoffruitand level
adultwomen vegetables
Havas[66] RCT USA Dietarycounselling 5 %changeinfruitand Education ↑*
intervention vegetablesconsumed level
Holme[57] RCT Norway Dietarycounselling 5 %changeincholesterol Socialclass ↑
intervention
Jeffery[67] RCT USA Healtheducation: 3 Meanweightchangein Household ↔*
Communitybased women(lb) income
education
Healtheducation: ↔*
Communitybased
educationwithan
additionalprizelottery
Jouret[58] RCT France Healtheducation:Healthy 4 Changein%ofchildren Arealevel ↓*
nutritionprogramaimedat overweight deprivation
children
Lowe[59] Cohortstudy UK Healtheducation:Healthy 3 %changeinvegetables Freeschool ↑
nutritionprogramaimedat observedconsumed meal
children entitlement
Plachta- RCT Germany Healtheducation:Healthy 5 Changein%prevalenceof Parental ↑*
Danielzik[60] nutritionprogramaimedat overweight education
children level
Reynolds[68] RCT USA Healtheducation:Healthy 3 Portionsoffruitand Household ↑*
nutritionprogramaimedat vegetablesconsumed income
children