Bond UniversityResearch Repository

Patient preferences for cardiovascular preventive medicationAlbarqouni, Loai; Doust, Jenny; Glasziou, Paul

Published in:Heart

DOI:10.1136/heartjnl-2017-311244

Published: 01/10/2017

Document Version:Peer reviewed version

Link to publication in Bond University research repository.

Recommended citation(APA):Albarqouni, L., Doust, J., & Glasziou, P. (2017). Patient preferences for cardiovascular preventive medication: Asystematic review. Heart, 103(20), 1578-1586. https://doi.org/10.1136/heartjnl-2017-311244

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

For more information, or if you believe that this document breaches copyright, please contact the Bond University research repositorycoordinator.

Download date: 10 Jul 2020

Patient preferences for cardiovascular preventive medication: a systematic review

Loai Albarqouni1, Jenny Doust1, Paul Glasziou1

1 Centre for Research in Evidence-Based Practice (CREBP), Bond University, QLD 4226, Australia

Loai Albarqouni: MD, MSc, PhD candidate; Jenny Doust: FRACGP, PhD, GP, and Professor of clinical epidemiology; Paul Glasziou: FRACGP, PhD, Director, and Professor of Evidence-Based Medicine

Corresponding author:

Loai Albarqouni, MD, MSc Centre for Research in Evidence-Based Practice (CREBP) Faculty of Health Sciences and Medicine Bond University QLD, Australia 4229 Email: loai.albaqrqouni@student.bond.edu.au Phone: +61(0)457699656

Manuscript word count: 3172 Abstract word count: 244 Number of tables and figures: 1 Table and 5 Figures Number of references: 50 Supplement: 1 Figure and 2 Tables

2

Abstract

Objective:Tosystematicallyreviewcurrentevidenceregardingtheminimumacceptablerisk

reductionofacardiovasculareventwhichpatientsfeelwouldjustifydailyintakeofa

preventivemedication.

Methods:WeusedtheWebofSciencetotracktheforwardandbackwardscitationsofasetof

fivekeyarticlesuntil15November2016.Studieswereeligibleiftheyquantitativelyassessed

theminimumacceptablebenefit-inabsolutevalues-ofacardiovasculardiseasepreventive

medicationamongasampleofthegeneralpopulation,andrequiredparticipantstochooseif

theywouldconsidertakingthemedication.

Results:Of253studiesscreened,weincluded22,involvingatotalof17,751participants:6

studiedProlongationofLife(POL),12AbsoluteRiskReduction(ARR),and14studiedNumber

NeededtoTreat(NNT)asmeasuresofriskreductioncommunicatedtothepatients.Instudies

framedusingPOL,39-54%(average48%)ofparticipantswouldconsidertakingamedicationif

itprolongedlife<8months,and56-73%(average64%)≥8months.Instudiesframedusing

ARR,42-72%(average54%)ofparticipantswouldconsidertakingamedicationthatreduces

their5-yearCVriskby<3%,and50-89%(average77%)≥3%.Instudiesframedusing5-year

NNT,31-81%(average60%)ofparticipantswouldconsidertakingamedicationwithaNNTof>

30,and46-87%(average71%)£30.

Conclusions:Manypatientsrequireasubstantialrisk-reductionbeforetheyconsidertakinga

dailymedicationworthwhile,evenwhenthemedicationisdescribedasbeingside-effectfree

andcostless.

Keywords:CardiovascularDiseases,Communication,Decision-making,RiskAssessment,GuidelineAbbreviation:CVDCardiovascularDisease;ARRAbsoluteRiskReduction;NNTNumberNeededtoTreat;POL

ProlongationofLife

3

Introduction

Cardiovasculardisease(CVD)istheleadingcauseofprematuredeathandreduceddisability

adjustedlife-yearsworldwide12.Thus,CVDprevention-whichmayrequiretheinitiationof

lifelongmedicationaimingatmodifyingspecificriskfactorsoffutureadverseevents3-has

been,andwillcontinuetobe,apublichealthpriority.

ClinicalPracticeGuidelines(CPGs)weredevelopedtoassistpractitionersandpatientsintheir

decisionsaboutappropriatehealthcareinterventions.However,diseasepreventiveguidelines

faceacrucialchallenge:individualsatriskoftenfeelhealthyandmightperceivemedicationsas

unnecessary,particularlythosehavinguncertainbenefitsand/orunpleasantsideeffects.This

challengecanleadtodisagreementbetweenhowguidelinespanelandpatientsvaluethe

benefitsandharmsofpreventivemedications4.Forinstance,whileguidelinespanelsmay

assumethatthebenefitsofapreventiveinterventionoutweighanypotentialharm,individual

patientsmaynotagreewiththeirtrade-off.Whenthereisuncertaintyinthetrade-off

betweenbenefitsandharmsofpreventiveinterventions5-7,incorporationofpatients’

preferencesisappropriateandimportant8.Accordingly,patientsmaydecidewhethertoaccept

orrejectsuchaninterventionbasedonweighingtheharms,cost,andinconvenienceoftaking

lifelongmedicationagainstthepotentiallong-termbenefits910.

Despitethegrowingacknowledgementoftheimportanceofpatientandpublicinvolvementin

thedevelopmentofCPGs,currentCPGsfocusprimarilyonthemedicationeffectivenessand

oftenfailtohighlightuncertaintyandreconcilepatientspreferencesandvalueswiththe

guidelinerecommendations11.Inaddition,CPGs’recommendationsareoftenusedasa

measureofthequalityofcareprovidedbycliniciansandorganisations1213.

Forinstance,themostrecentupdateoftheUSandUKCVDriskassessmentandprevention

guidelineshaverecommendedtheuseofstatinsamongpeoplepreviouslyconsideredatlow

4

riskofcardiovasculardisease,settingthethresholdtocommencestatinsatariskofaCVD

eventin10yearsof7.5%1415.However,thisthresholdreflectsavaluejudgmentaboutthe

balanceofpotentialharmsandbenefitsoftakingstatins,whichsomepatientsmaynotfindit

justifiablefortakingstatins.

Therefore,thepurposeofthissystematicreviewwastosummarisecurrentevidenceregarding

theminimumacceptableriskreductionofacardiovasculareventwhichpatientsfeelwould

justifytheirtakingdailypreventivemedication.Thisinformationishelpfulbothforpatients,to

activelyparticipateinthedecision-makingprocess16,andguidelinespanelstobeinformedof

thebestevidenceregardingthepatientsvaluesandpreferencesinmakingtrade-offsbetween

desirableandundesirableconsequencesofsuchanintervention17.

Methods

Aprotocolforthissystematicreviewwasdevelopedinadvance(availablefromtheauthorson

request).

InformationSourcesandSearchMethods

Wedeveloped,withthehelpofamedicallibrarianexperiencedinsystematicreviews,asearch

strategyusingamethodofforwardandbackwardscitationanalysis.WeusedWebofScience

databasetotracktheforwardandbackwardscitationsofasetof5keyarticlesuntil15-

November-2016.Theindexarticlesforourcitationanalysiswerearelevantsystematicreview18

andfourotherarticlesidentifiedasimportantinthisarea9101920.Searchingcontinued

backwardandforwarduntilnofurtherrelevantstudiesarefound.Citationanalysiscanevade

thetime-consumingandthecomplexnatureofthestandardsearchstrategies,inareaswhere

indexingisunlikelytoretrieverelevantarticleswithanacceptableaccuracyrate21-23.

5

EligibilityCriteria

Allpublishedquantitativeprimarystudieswereeligible,irrespectiveofthestudydesign.We

includedstudiesinvolvingadultparticipantswhomadeadecisionaboutacardiovascular

preventivemedicationwhetherinareal-life(actual)orahypothetical(analogous)scenario.

Studieswereeligibleiftheyassessedtheminimumacceptablebenefit(CVD-risk-reduction)ofa

CVDpreventivemedicationamongasampleofthegeneralpopulation,andprovideda

quantitativeestimateoftheriskreduction(inabsolutevalues)requiredbypatientstomake

dailymedicationworthwhile.Absoluteriskreductionestimatescouldbepresentedinthe

formatofProlongation-of-Life(POL),Absolute-Risk-Reduction(ARR)orNumber-Needed-to-

Treat(NNT).Studiesthatprovidedadescriptiveorqualitativeestimateofriskreductionora

quantitativeestimateofrelative-risk-reduction(withoutbaseline-risk)wereexcluded.

SelectionofStudies

Twoauthorsindependentlyscreenedforeligibilitythetitlesandabstractsofidentifiedarticles.

Weretrievedthefull-textofstudiesthatpotentiallymettheeligibilitycriteria.Fromthefull-

texts,twoauthorsindependentlyassessedstudyeligibility.Whenmorethanonepublicationof

thesamestudyexisted,thepublicationwiththemostcompletedatawasincluded.We

resolvedanydisagreementbydiscussion.Reasonsforexclusionofstudiesweredocumented.

DataExtraction

Twoauthorsindependentlyabstracteddata,usingastandardiseddataextractionform.The

abstracteddataincluded:(1)General:title,authors,country,languageandyearofpublication,

duplicatepublications,sponsoring;(2)Participants:samplesize,baselinecharacteristics,and

studysetting;(3)Scenario:Hypothetical/Real,typeofriskpresentationformat(POL,ARRor

NNT),methodforelicitingpatientspreferences(e.g.discretechoice),medication

6

characteristics(sideeffects,cost,burden)andthetargetadverseevent(mortalityor

morbidity).

AssessmentofthemethodologicalqualityandRiskofBias

Includedstudieswereindependentlyassessedbytwoauthorsusingamodifiedversionofthe

riskofbiasinprevalencestudiesassessmenttool24,whichincludesthefollowingitems:

representation,selection,responserate,datacollection.Weassessedtheadequacyofeach

item:'low',‘unclearor‘high’riskofbias.

DataSynthesis

Foreachincludedstudy,weextractedorcalculatedtheaverage5-yearARRorNNTorPOL

requiredbytheparticipantstocommenceaCVDpreventivemedication.Wereportedthe

ranges(andsample-size-weightedaverages)of5-yearARRorNNTorPOLacrossincluded

studies.Ameta-analysiswasnotpossibleduetotheheterogeneityinstudies’methodsand

outcomes.

Results

Wescreened341studies(afterremovalofduplicates),ofwhich238wereclassedasineligible

basedontitlesorabstracts.Oftheremaining103full-texts,22studieswithatotalof17751

participantswereincluded(Figure1).

Studypublicationdatesrangedfrom1995to2014;studysample-sizerangedfrom58to2978,

withanaverageof807participants.Threestudieswereinternet-based(1US,1Norway,and1

USandNorway),whiletheremainingstudieswereconductedinsevencountries:6UK,5

Denmark,2Norway,2NewZealand,2US,and2Canada.The22includedstudiescontributed

63estimatesofminimumacceptableriskreduction:14studiescontributed28estimatesinthe

formofnumber-needed-to-treat(NNT);12studiescontributed22estimatesintheformof

absolute-risk-reduction(ARR)and6studiescontributed13estimatesintheformof

7

prolongation-of-life(POL).Participantsweresampledfromthegeneralpopulationin13studies

(59%)andfromprimary-caresettingsin9studies(41%).Moststudieselicitedparticipants’

preferencesusingsingle-choicequestions(n=16;70%).Othermethodsweremultiple-choice

questions(n=4),iterativeprocess(n=1)ortrade-offmethod(n=1).Themostfrequentmethod

ofdatacollectionwasin-personinterviews(n=12;eithertelephoneorface-to-face);the

remainingstudiesusedsurveystocollectdatafromtheparticipants(eithermailedorinternet-

basedsurvey).CharacteristicsofincludedstudiesareshowninTable1.

Theresultsofthequalityassessment(Figure2)showedthathalfofthestudiestruly

representedthegeneralpopulation(n=11).Fewstudies(n=3)hadaresponserate≥75%,but

randomselectionorcensushadbeenundertakeninmoststudies(n=17).Datahadbeen

collectedusingthesamemethodsfromallparticipantsinmostincludedstudies(n=17).

ProlongationofLife(POL)

InstudieswhichpresentedtheCVbenefitsframedasPOL,39-73%oftheparticipants(average

54%)wouldconsidertakingcardio-preventivemedication.Thisdecreasedto39-54%(average

48%)whenweonlyconsideredpreventivemedicationsthatprolongedtheirlifebylessthan8

months,andincreasedto56-73%(average64%)whentheanalysiswasrestrictedtostudies

presentedPOLequaltoormorethan8(Figure3andeTable1intheSupplement).

Absoluteriskreduction(5-year)

InstudieswhichpresentedtheCVbenefitsintermofARR,42-89%oftheparticipants(average

60%)wouldconsidertakingtheCVDpreventivemedication.Thisdecreasedto42-72%

(average54%)whenweonlyconsiderstudieswheremedicationreducedthe5-yearabsolute

CVriskbylessthan3%,whilethispercentageincreasedto50-89%(average77%)whenwe

considerstudiespresentedARRequaltoormorethan3%(Figure4andeTable2inthe

Supplement).

8

NumberNeededtoTreat(5-year)

InstudieswhichpresentedtheCVbenefitsintermofNNT,31-87%oftheparticipants(average

64%)wouldconsidertakingcardio-preventivemedications.Thisdecreasedto31-81%(average

60%)whenweonlyconsideredstudiespresentedamedicationneededtobetakenbymore

than30personstopreventoneeventwhilethispercentageincreasedto46-87%(average

71%)whenanalysisrestrictedtostudiespresentedNNTequaltoorlessthan30(Figure5and

eTable2intheSupplement).

FactorsaffectingthedecisiontocommenceaCVDpreventivetherapy

Somestudiesexploredotherfactorsthatwouldaffectthedecisiontotakeamedication.These

factorshelptoexplainsomeoftheheterogeneityobservedbetweenstudies.

Probabilisticvs.deterministic‘fixed’presentationofrisk

OnestudyfoundthatparticipantsweremorelikelytoagreetoaCVDpreventivetherapywhen

thebenefitswerepresentedasacertain,short,postponementofaheartattack(182/456;

40%)comparedtoanequalaveragebutuncertain‘probabilistic’longpostponementofthe

outcome(153/452;34%)19.ThiscognitivebiashasbeenalsoshownbyFinegoldetal25.

Participantspresentedwithacertain1-yearlifegainweremorelikelytoacceptamedication

thanwhenpresentedwitha10%chanceof10-yearlifegain(56%vs45%)25.However,if

participantswerepresentedwithtwogainsthatwerenotprobabilisticallyequivalent,they

preferredthetherapythatprovidesthelargergainwhetherpresentedasacertainoruncertain

benefit25.

PositiveversusNegativeframing

Carlingetal26foundthatparticipantsweremorelikelytoagreeforaCVDmedicationwhenthe

riskreductionwasframednegatively(66%)thanpositively(56%)orcomparedtoinformed

informationthatincludedpositiveandnegativeformattogether(40%and46%respectively).

9

Similarly,Goodyear-Smith27foundthatparticipantsweremorelikelytoagreetotakingaCVD

medicationwhenthebenefitswereframednegatively(89%)thanpositively(80%).

Riskpresentationformat

Stovringetal28foundthattheproportionofrespondentswillingtotakeamedicationto

preventCVDwashigherwhenthebenefitwaspresentedaseitherARR(expressedinnatural

frequencies)orNNTcomparedtoPOL.Notably,POLwasfoundtohavetheleastconcordance

withfinaldecisionsthatwerebasedoncomprehensiveinformation(ARR,NNT,RRRand

pictorialpresentation).However,anotherstudyfoundmorepatientsconsideredtaking

medicationwhentheresultswerepresentedasARR(85%)thanNNT(67%)27.Carlingetal29

compareddifferentpresentationsofriskreduction,andfoundthatparticipantsweremore

likelytoconsideramedicationwhenthebenefitpresentedasrelativeriskreduction,but

naturalfrequencieswerescoredthehighestinpreference,understandingandsatisfactionwith

thepresentedinformation,andconfidenceinthedecision.

Addingadverseeffects

Friedetal30askedcommunity-livingolderpersonstheirwillingnesstoagreetoCVDpreventive

medicationswithvaryingdegreeofbenefitsandharms(bothtypeandseverity),andfound

thataboutahalfoftherespondentswhoinitiallyagreedtoamedicationwithaveragebenefit

withnoadverseeffectswereunwillingoruncertainabouttakingthesamemedicationifitwas

associatedwithevenmildfatigueandnausea,andonly3%ofthemwouldagreetoa

medicationthathadadverseeffectsthatcouldaffecttheiractivitiesofdailyliving.

PrevioushistoryorhighriskforCVD

ThepresenceofknownriskfactorsorahistoryofCVDmightalsoaffectthedecisiontotake

themedication.Forinstance,MisselbrookandArmstrong31foundthatpatientswithknown

highbloodpressureweremorelikelytoagreetoaCVDpreventivemedicationthan

10

participantswithnormalbloodpressure,whichhasbeenalsofoundbyMcAlisteretal32.

SimilarfindingswerealsoshownbyTrewbyetal33comparingparticipantswithahistoryofCVD

andtakingmedicationstopatientseitherwithoutCVDhistorybuttakingapreventive

medicationorwithoutCVDhistoryandtakingnomedication.Similarly,Dahletalfoundthat

previoushistoryofmyocardialinfarction(OR:2.595%CI1.3-5.0),andelevatedcholesterollevel

(OR:1.995%CI1.8-2.6)significantlyincreasedtheoddsofconsentingtomedication34.Friedet

alfoundthatparticipantswhoratedtheirhealthaspoor/fairweremorelikelytoconsider

takingamedicationthanthoseratedtheirhealthasexcellent/verygood30.However,twoother

studiesfoundnoassociationbetweenself-reportedhealth,CVDhistoryandtheoddsof

consideringtakingmedication1935.

Participantbaselinecharacteristics

Malegenderwasfoundtobesignificantlyassociatedwithconsideringtakingamedicationin

Dahletal34,Halvorsenetal(2005)36,andHudsonetal3,but,notinHalvorsenetal(2007)19,

Nexoeetal37,andKristiansenetal35.Olderageandhighereducationlevelwassignificantly

associatedwithconsideringtakingamedicationinHalvorsenetal(2007)19,andKristiansenet

al35,but,notinNexoeetal37,Halvorsenetal(2005)36,andDahletal34.

PerspectiveofPatientsvs.Clinicians

McAlisteretal32comparedtheminimumgaininARRtoconsidertakingCVDmedication

betweenhypertensivepatients(withnoovertCVD)andfamilyphysicians,andfoundthat

patientsdemandgreaterbenefitsbeforeagreeingtoaCVDpreventivemedicationcompared

tophysicians.Similarly,Steel38foundthatthegaininARRtoagreetoanantihypertensive

medicationwasloweramongconsultantphysiciansthangeneralpractitioners,andhighest

amongthegeneralpublic.

11

Discussion

Resultsofthe22studiesinthissystematicreviewsuggestthatanimplicitassumptionin

cardiovascularpreventionguidelines-thatparticipantsaccepttheamountbalancebetween

benefitsandharms-maynotbeconcordwithpatientpreferences.Eveninthestudieswhere

peoplewerepresentedwithanidealisedtabletwithnosideeffects,morethanone-third

statedtheywouldnotconsidertakingamedicationthatcouldpreventtheir5-yearabsolute

riskofacardiovasculareventbyanARRof5%ormore.ThisgainisrarelyachievedbyCVD

medication(e.g.statinreducetherelativeriskofaCVDeventbyabout5-yearCVabsoluterisk

byabout30%3940.Moreover,considerablevariationswereobservedinthepatients’

preferencesoftheminimumacceptableriskreductionofcardio-preventivemedication.

Severalfactorsmayexplainthewidevariationinpatients’preferencesbothwithinandacross

includedstudies,suchaswhetherpatientsarealreadytakingapreventivemedication,andthe

waythattheriskreductionhasbeencommunicatedtotheparticipants(e.g.riskpresentation

formatandframingeffect).Arecentsystematicreviewofriskcommunicationfoundthatthe

variousriskpresentationformatsmighthavedifferenteffectsonthedecisiontoconsider

takingmedication41.However,itisimportanttodistinguishbetweenthepersuasivenessofthe

riskpresentationformatsandtheclarityandunderstandabilityoftheformat18.Methodsused

toelicitpreferences(e.g.single-choiceoptions,trade-off)aswellasthoseusedtocollectdata

(face-to-faceinterviews,mailedquestionnaire)mayalsocontributetovariationobserved.

Lastly,thecharacteristicsofthemedicationpresentedtoparticipantsmayalsohaveaffected

theirdecisions.Anidealmedication(freeofchargeandnosideeffects)ismorelikelytobe

acceptedcomparedtoareal-lifemedicationwithcostsandpotentialsideeffects.

Thus,whenpatients’preferencesvarywidely,preventiveguidelinesshouldbewaryofsettinga

singleriskthresholdfortheinitiationofapreventivetreatment,basedontheeffectivenessof

12

themedication.Anythresholdpointorrangeshouldbebasedonthebalanceofbenefitsand

downsides,whichshouldbeinformedbypatients’valueandpreferences.

StrengthandLimitation

Theheterogeneityofmethodsusedtoassessthepreferences,includingtrade-offmethod,

iterativeprocess,andsinglechoicequestionsisbothastrengthandlimitationofourstudy.The

inclusioncriteriameantthatawidevarietyofestimatescouldbeincludedinthereview,but

owingtothisheterogeneityofthemethodsandoutcomes,wecouldnotcalculatequantitative

summaryestimatesoftheminimumacceptableriskreduction,whichiscommoninsystematic

reviewsinthisfield174243.Moreover,mostoftheincludedstudieshavenotbeendesigned

primarilytoassessourresearchquestion,buttoexamineotherissuessuchasresponseto

framing.Inaddition,thereviewonlyconsidersthosefactorsreportedintheincludedstudies.

Anotherpotentiallimitationistheriskofmissingrelevantarticlesbecauseoftheuseofcitation

analysisasasearchstrategy.However,studiesinotherareassuggesttheaccuracyrateofthis

methodisacceptable(e.g.usingcitationanalysis,JanssensandGwinn retrievedabout94%

[75-100%]ofallincludedarticlesretrievedusingtraditionalmethodin10meta-analyses23).

Thus,wethinkthatouroverallresultsandconclusionareunlikelytobeaffected.

Implicationforclinicalpractice

Patients’preferencesshouldbeconsideredinbothguidelinedevelopmentandclinicalcare.

Patientsrepresentativesandmemberofthepublicshouldbeactivelyinvolvedinguideline

developmentpanels44.Futureguidelinesmustconsidermorethantheevidenceofmedication

effectivenessinestablishingtreatmentthresholds:Systematicreviewsofpatients’valuesand

preferencesshouldalsobeconductedtoinformguidelinerecommendations.Giventhe

heterogeneityinindividualpreferences,guidelinesshouldincludestrategiestohelpclinicians

incorporatepatients’preferencesandvaluestomakeinformeddecisions.Thiscanbeachieved

13

bypromotingchoice,providingadequateinformationaboutthebenefitandrisksof

medication,andthelevelofuncertaintyintheevidence,andhighlightingthatguideline

recommendationssupport,butdonotsubstituteforclinicaljudgmentandindividualised

decision-making.Visualaids(e.g.iconarraysorriskpictograph)tocommunicateriskreduction

mightbeusedsinceevidenceshowsthatsuchaidsimprovepatients’understandingand

satisfaction41.Thisalsoimpliesthatthe‘doordonot’styleofrecommendationsshould

generallybediscouraged,andthatpreference-sensitiverecommendationsshouldnotbeused

asindicatorsofqualityofcare(qualitymeasures).

Cliniciansoftenfailtoaccuratelyidentifypatientpreferences–aproblemtermedsilent

misdiagnosis45,andhenceguidelinesshouldincludetoolsthatpromoteshareddecisionmaking

whereinterventionschoicescanbetailorednotonlytopatients’clinicalcharacteristicsbutalso

totheirvaluesandpreferences.Thismeansthatguidelinerecommendationsshouldbeusedas

astartingpointfortheindividualpatient-cliniciandecision-makingprocess.Inaddition,patient

versionsofguidelines,aswellasdecisionaidstools,shouldbedevelopedanddisseminated

alongwiththeguidelines.

Implicationforfutureresearch

Thissystematicreviewhighlightstheneedforstandardisingthekeymeasuresofpatient

preferencesthatmightbeusedtoinformguidelines.Futureresearchcouldgainfromusing

well-establishedmethodsforpreferenceelicitationusedinotherdisciplines(e.g.health

economics)todevelopavalid,reliable,andsensitivemeasureofpatientpreferencesina

medicaldecision-makingcontext.Moreover,asystematicreviewofqualitativestudies

regardingthefactorsaffectstheindividualdecisionofCVpreventivemedicationisalso

warranted.

14

Conclusion

Evenforaside-effectfree,costlessmedication,manypatientsrequireasubstantialrisk-

reductiontoconsidertakingapreventivemedication.However,therangeofanswersisvery

widewithsomeacceptinganygain,andothernotwillingtoconsidermedicationwithanygain.

Guidelinesandclinicalconsultationsneedtoaccountforboththeseaverageandindividual

valuesinsettingriskthresholds.

ContributorsLA,JD,PGdesignedthestudy.LA,JD,PGscreenedthearticles.LA,JD,PGextracteddata.LAdraftedtheoriginalmanuscriptandallauthorsrevisedthepaper.LAistheguarantorofthestudy.

AcknowledgmentWewouldliketothankMsSarahThorningforherhelpwiththeliteraturesearch.

CompetinginterestsAllauthorshavecompletedtheICMJEuniformdisclosureformatwww.icmje.org/coi_disclosure.pdfanddeclare:DrGlaszioureportsgrantsfromNationalHeartFoundationpriortotheconductofthisstudy.Nosupportfromanyorganisationforthesubmittedwork;nootherrelationshipsoractivitiesthatcouldappeartohaveinfluencedthesubmittedwork.

FundingNospecificfundingforthisresearch

EthicalapprovalNotrequired.

DatasharingDataextractedfromtheincludedstudiesareavailableonrequestfromthecorresponding

author.

TheCorrespondingAuthorhastherighttograntonbehalfofallauthorsanddoesgrantonbehalfofall

authors,anexclusivelicence(ornonexclusiveforgovernmentemployees)onaworldwidebasistothe

BMJPublishingGroupLtdanditsLicenseestopermitthisarticle(ifaccepted)tobepublishedinHEART

editionsandanyotherBMJPGLproductstoexploitallsubsidiaryrights

15

References1.MurrayCJ,LopezAD.Measuringtheglobalburdenofdisease.NEnglJMed2013;369(5):448-57.

2.MortalityGBD,CausesofDeathC.Global,regional,andnationalage-sexspecificall-causeandcause-specificmortalityfor240causesofdeath,1990-2013:asystematicanalysisfortheGlobalBurdenofDiseaseStudy2013.Lancet2015;385(9963):117-71.

3.HudsonB,ZarifehA,YoungL,etal.Patients'expectationsofscreeningandpreventivetreatments.AnnFamMed2012;10(6):495-502.

4.HarmsenCG,StovringH,JarbolDE,etal.Medicationeffectivenessmaynotbethemajorreasonforacceptingcardiovascularpreventivemedication:apopulation-basedsurvey.BMCMedInformDecisMak2012;12:89.

5.AbramsonJD,RosenbergHG,JewellN,etal.Shouldpeopleatlowriskofcardiovasculardiseasetakeastatin?BMJ2013;347:f6123.

6.MartinSA,BoucherM,WrightJM,etal.Mildhypertensioninpeopleatlowrisk.BMJ2014;349:g5432.

7.RedbergRF,KatzMH.Statinsforprimaryprevention:Thedebateisintense,butthedataareweak.JAMAInternalMedicine2016.

8.HoffmannTC,LegareF,SimmonsMB,etal.Shareddecisionmaking:whatdocliniciansneedtoknowandwhyshouldtheybother?MedJAust2014;201(1):35-9.

9.HuxJE,NaylorCD.Communicatingthebenefitsofchronicpreventivetherapy:doestheformatofefficacydatadeterminepatients'acceptanceoftreatment?MedDecisMaking1995;15(2):152-7.

10.FontanaM,AsariaP,MoraldoM,etal.Patient-accessibletoolforshareddecisionmakingincardiovascularprimaryprevention:balancinglongevitybenefitsagainstmedicationdisutility.Circulation2014;129(24):2539-46.

11.SchunemannHJ,FretheimA,OxmanAD.Improvingtheuseofresearchevidenceinguidelinedevelopment:10.Integratingvaluesandconsumerinvolvement.HealthResPolicySyst2006;4:22.

12.WalterLC,DavidowitzNP,HeinekenPA,etal.Pitfallsofconvertingpracticeguidelinesintoqualitymeasures:lessonslearnedfromaVAperformancemeasure.JAMA2004;291(20):2466-70.

13.GarberAM.Evidence-basedguidelinesasafoundationforperformanceincentives.HealthAff(Millwood)2005;24(1):174-9.

14.NationalInstituteforHealthAndCareExcellence(NICE).Cardiovasculardisease:riskassessmentandreduction,includinglipidmodification.2014.

15.GoffDC,Jr.,Lloyd-JonesDM,BennettG,etal.2013ACC/AHAguidelineontheassessmentofcardiovascularrisk:areportoftheAmericanCollegeofCardiology/AmericanHeartAssociationTaskForceonPracticeGuidelines.Circulation2014;129(25Suppl2):S49-73.

16.OsterbergL,BlaschkeT.Adherencetomedication.NEnglJMed2005;353(5):487-97.

17.MacLeanS,MullaS,AklEA,etal.Patientvaluesandpreferencesindecisionmakingforantithrombotictherapy:asystematicreview:AntithromboticTherapyandPreventionofThrombosis,9thed:AmericanCollegeofChestPhysiciansEvidence-BasedClinicalPracticeGuidelines.Chest2012;141(2Suppl):e1S-23S.

18.WaldronCA,vanderWeijdenT,LudtS,etal.Whatareeffectivestrategiestocommunicatecardiovascularriskinformationtopatients?Asystematicreview.PatientEducCouns2011;82(2):169-81.

16

19.HalvorsenPA,SelmerR,KristiansenIS.Differentwaystodescribethebenefitsofrisk-reducingtreatments-Arandomizedtrial.AnnInternMed2007;146(12):848-56.

20.HarmsenCG,KristiansenIS,LarsenPV,etal.Communicatingriskusingabsoluteriskreductionorprolongationoflifeformats:cluster-randomisedtrialingeneralpractice.BrJGenPract2014;64(621):E199-E207.

21.RobinsonKA,DunnAG,TsafnatG,etal.Citationnetworksofrelatedtrialsareoftendisconnected:implicationsforbidirectionalcitationsearches.JClinEpidemiol2014;67(7):793-9.

22.BelterCW.Citationanalysisasaliteraturesearchmethodforsystematicreviews.JournaloftheAssociationforInformationScienceandTechnology2016;67(11):2766-77.

23.JanssensAC,GwinnM.Novelcitation-basedsearchmethodforscientificliterature:applicationtometa-analyses.BMCMedResMethodol2015;15:84.

24.HoyD,BrooksP,WoolfA,etal.Assessingriskofbiasinprevalencestudies:modificationofanexistingtoolandevidenceofinterrateragreement.JClinEpidemiol2012;65(9):934-9.

25.FinegoldJA,Shun-ShinMJ,ColeGD,etal.Distributionoflifespangainfromprimarypreventionintervention.OpenHeart2016;3(1):e000343.

26.CarlingCL,KristoffersenDT,OxmanAD,etal.Theeffectofhowoutcomesareframedondecisionsaboutwhethertotakeantihypertensivemedication:arandomizedtrial.PLoSOne2010;5(3):e9469.

27.Goodyear-SmithF,ArrollB,ChanL,etal.Patientspreferpicturestonumberstoexpresscardiovascularbenefitfromtreatment.AnnFamMed2008;6:213-17.

28.StovringH,Gyrd-HansenD,KristiansenIS,etal.Communicatingeffectivenessofinterventionforchronicdiseases:whatsingleformatcanreplacecomprehensiveinformation?BMCMedInformDecisMak2008;8:25.

29.CarlingCLL,KristoffersenDT,MontoriVM,etal.TheEffectofAlternativeSummaryStatisticsforCommunicatingRiskReductiononDecisionsaboutTakingStatins:ARandomizedTrial.PLoSMed2009;6.

30.FriedTR,TinettiME,TowleV,etal.Effectsofbenefitsandharmsonolderpersons'willingnesstotakemedicationforprimarycardiovascularprevention.ArchInternMed2011;171(10):923-8.

31.MisselbrookD,ArmstrongD.Patients'responsestoriskinformationaboutthebenefitsoftreatinghypertension.BrJGenPract2001;51:276-79.

32.McAlisterFA,O'ConnorAM,WellsG,etal.Whenshouldhypertensionbetreated?ThedifferentperspectivesofCanadianfamilyphysiciansandpatients.CMAJ2000;163(4):403-8.

33.TrewbyPN,ReddyAV,TrewbyCS,etal.Arepreventivedrugspreventiveenough?Astudyofpatients'expectationofbenefitfrompreventivedrugs.ClinMed(NorthfieldIl)2002;2:527-33.

34.DahlR,Gyrd-HansenD,KristiansenIS,etal.Canpostponementofanadverseoutcomebeusedtopresentriskreductionstoalayaudience?Apopulationsurvey.BMCMedInformDecisMak2007;7.

35.KristiansenIS,Gyrd-HansenD,NexoeJ,etal.Numberneededtotreat:easilyunderstoodandintuitivelymeaningful?Theoreticalconsiderationsandarandomizedtrial.JClinEpidemiol2002;55:888-92.

36.HalvorsenPA,KristiansenIS.Decisionsondrugtherapiesbynumbersneededtotreat-Arandomizedtrial.ArchInternMed2005;165:1140-46.

37.NexoeJ,KristiansenIS,Gyrd-HansenD,etal.Influenceofnumberneededtotreatcostsandoutcomeonpreferencesforapreventivedrug.FamPract2005;22:126-31.

38.SteelN.Thresholdsfortakingantihypertensivedrugsindifferentprofessionalandlaygroups:questionnairesurvey.BMJ2000;320(7247):1446-7.

17

39.ChouR,DanaT,BlazinaI,etal.StatinsforPreventionofCardiovascularDiseaseinAdults:EvidenceReportandSystematicReviewfortheUSPreventiveServicesTaskForce.JAMA2016;316(19):2008-24.

40.CollinsR,ReithC,EmbersonJ,etal.Interpretationoftheevidencefortheefficacyandsafetyofstatintherapy.Lancet2016.

41.ZipkinDA,UmscheidCA,KeatingNL,etal.Evidence-basedriskcommunication:asystematicreview.AnnInternMed2014;161(4):270-80.

42.HoffmannTC,DelMarC.Patients'expectationsofthebenefitsandharmsoftreatments,screening,andtests:asystematicreview.JAMAInternMed2015;175(2):274-86.

43.EiringO,LandmarkBF,AasE,etal.Whatmatterstopatients?Asystematicreviewofpreferencesformedication-associatedoutcomesinmentaldisorders.BMJOpen2015;5(4):e007848.

44.BoivinA,CurrieK,FerversB,etal.Patientandpublicinvolvementinclinicalguidelines:internationalexperiencesandfutureperspectives.QualSafHealthCare2010;19(5):e22.

45.MulleyAG,TrimbleC,ElwynG.Stopthesilentmisdiagnosis:patients'preferencesmatter.BMJ2012;345:e6572.

46.CarlingC,KristoffersenDT,HerrinJ,etal.Howshouldtheimpactofdifferentpresentationsoftreatmenteffectsonpatientchoicebeevaluated?Apilotrandomizedtrial.PLoSOne2008;3(11):e3693.

47.LeamanH,JacksonPR.Whatbenefitdopatientsexpectfromaddingsecondandthirdantihypertensivedrugs?BrJClinPharmacol2002;53:93-99.

48.MarshallT,BryanS,GillP,etal.Predictorsofpatients'preferencesfortreatmentstopreventheartdisease.Heart2006;92(11):1651-5.

49.NicholsonK,RamsayLE,HaqIU,etal.Factorsaffectingtheacceptanceofdrugtherapytopreventmyocardialinfarction.BrJClinPharmacol1999;47(5):580-80.

50.SorensenL,Gyrd-HansenD,KristiansenIS,etal.Laypersons'understandingofrelativeriskreductions:Randomisedcross-sectionalstudy.BMCMedInformDecisMak2008;8.

18

Figures

Figure1:Summaryofflowofinformationthroughthesystematicreviewprocess(PRISMAFlowDiagram)

19

Figure2:Riskofbiasassessment:authors’judgmentsabouteachriskofbiasitem:(A)foreachincludedstudy(Riskofbiasgraph);and(B)presentedaspercentagesacrossallincludedstudies(riskofbiassummary).Representation:wasthesamplingframeatrueorcloserepresentativeofthetargetpopulation/generalpopulation;Selection:

wassomeformofrandomselectionused oracensusundertaken;Non-response:wasthelikelihoodofnon-responsebiasminimal(lowrisk:≥75%,highrisk:<75%);Datacollection:weredatacollecteddirectlyfromparticipantsandifthesamemode

ofdatacollectionusedforallparticipants.

20

Figure3:Minimumacceptableriskreductionincardiovascularevents(expressedasProlongationof

Life)requiredbythepatientstojustifytakingdailypreventivemedication.Bubblegraphshowingthepercentageofparticipantsofincludedstudiesthatcouldaccepttakingamedicationagainstthe

potentialprolongationinlife(inmonthsinlogarithmicscale)forsuchamedication;thebubblearearepresentsthesample

size.

21

Figure4:Minimumacceptableriskreductionincardiovascularevents(expressedasARR)requiredby

thepatientstojustifytakingdailypreventivemedicationBubblegraphshowingthepercentageofparticipantsofincludedstudiesthatcouldaccepttakingamedicationagainstthe5-

yearabsoluteriskreductionincardiovasculareventsthatcanbegainedbysuchamedication;thebubblearearepresentsthe

samplesize.

22

Figure5:Minimumacceptableriskreductionincardiovascularevents(expressedasNNT)requiredby

thepatientstojustifytakingdailypreventivemedication.Bubblegraphshowingthepercentageofparticipantsofincludedstudiesthatcouldaccepttakingamedicationagainstthe

numberofindividualneededtobetreatedfor5yearstopreventonecardiovascularevent;thebubblearearepresentsthe

samplesize.

23

Tables

Table1.Characteristicsofincludedstudies

Study Participants: Sample Size

(response rate%); Age; Women (%); data collection

Participants Characteristics Settings

Scenario: risk; medication; adverse event; side effects;

cost Outcomes

Carling, 2008, Internet-based46

770 participants; 62% between 40-59 years; 58% women; internet-based survey (in English)

Community (18+; internet-users); 90% educated to more than 12 years; 84% from USA; 23% GP or health professionals;

Hypothetical; single choice experiment; high cholesterol level; daily statin-like; 10-year CV risk of angina or heart attack; side effects of statins; 50US$ per month

To decide whether to take the medication or not based on a single value of risk reduction (ARR, NNT)

Carling, 2009, Internet-based29

2978 participants; 54% between 40-59 years; 59% women; internet-based survey (in English and Norwegian)

Community (18+; internet-users); 91% educated to more than 12 years; 42% from USA

Hypothetical; single choice experiment; high cholesterol level; daily statin-like; 10-year CV risk of angina or heart attack; side effects of statins; 50US$ per month

To decide whether to take the medication or not based on a single value of risk reduction (ARR or NNT)

Carling, 2010, Internet-based26

1528 participants; 49% between 40-59 years; 53% women; internet-based survey (in Norwegian)

Community (18+; internet-users); 59% educated to university-level

Hypothetical; single choice experiment; 40-yr old, non-smoker, active and has a healthy diet; high blood pressure level; daily; 10-year CV risk of stroke or heart attack; dizziness, impotence nausea, muscle cramps and others; co-payment

To decide whether to take the medication or not based on a single value of risk reduction, framed both positively and negatively (ARR)

Dahl, 2007, Denmark34

1367 participants (50%); mean age 60; 52% women; in-person interview

Community (40+); 32% had elementary education only; 5% history of heart attack and 18% hypercholesterolemia

Hypothetical; single choice experiment; high cholesterol level; daily; heart attack; few and harmless side effects; 60€ per year

To decide whether to take the medication or not based on a single value of risk reduction (POL)

Fontana, 2014, UK10

360 participants; mean age 38 years; 50% women; face-to-face interview

Community; 22% on regular medications; 1% history of previous CVD

Hypothetical; trade-off; high cholesterol level; daily statin-like; CVD and death; no side effects; negligible

Medical aversion (gain in lifespan required by participants to commence lifelong therapy); result was extracted from Figure1

Fried, 2011, US30

356 participants; mean age 76; 75% women; in-person interview

Community (senior centres); mean years of education was 13; 69% had more than 3 chronic conditions; 92% took 1 or more medication, with a mean of 4

Hypothetical; single choice experiment; multiple scenarios; primary CVD prevention; daily; 5-year MI risk; no side effects; covered by insurance

To decide whether to take the medication or not based on a single value of risk reduction (ARR) + pictograph

Goodyear-Smith, 2008, New Zealand27

100 participants (53%); mean 66 years; 60% women; telephone interview

Outpatients (4 family practices); patients eligible if they had heart disease and on statin;

Hypothetical; single choice experiment; angina or heart attack; daily statin-like; 5-year heart attack risk; few side effects;

To decide whether to take the medication or not based on a single value of risk reduction (ARR or NNT); negative framing

Halvorsen, 2005, Norway36

1201 participants (60%); 43% between 25-44 year; 48% women; face-to-face (or telephone) interview

Community; 28% educated more than 12 years; 11% history of hypertension, 8% hypercholesterolemia

Hypothetical; multiple choice experiment; multiple scenarios; angina or heart attack; daily statin-like; heart

To decide whether to take the medication or not based on a single value of risk reduction (3-year NNT)

24

attack or stroke; no serious side effects; cost comparable to common cardio-preventive medication

Halvorsen, 2007, Norway19

1397 participants (81%); mean 58 years; 34% women; mailed questionnaire

Community; 23% educated more than 12 years; 21% previous CVD history

Hypothetical; single choice experiment; daily statin-like; heart attack; neither common nor dangerous side effects; refunded

To decide whether to take the medication or not based on a single value of risk reduction (5-year NNT, POL)

Hudson, 2012, New Zealand3

354 participants (36%); mean 60 years; 44% women; mailed questionnaire

Outpatients (3GPs, all registered 50-70 years); 31% educated to high school only; 13% previous CVD history

Hypothetical; single choice experiment; daily; death; no major side effects; 3$ per 3 months

To choose, from six different estimates (10-year NNT), the number of lives should be saved to justify lifelong medication

Hux, 1995, Canada9

100 participants; 53% older than 55 years; 47% women; questionnaire

Outpatients (35-65 years); 55% take cardiovascular medications; 70% history of heart disease; 62% attended cardiology clinic

Hypothetical; multiple choice experiment; daily statin-like; heart attack; negligible side effects; insured

To decide whether to take the medication or not based on a single value of risk reduction (5-year ARR, NNT, POL)

Kristiansen, 2002, Denmark35

675 participants (60%); mean age 44 years; 51% women; face-to-face interview

Community; 21% educated to elementary-level only; 3.3% previous CVD history

Hypothetical; single choice experiment; daily; heart attack; few and mild side effects; co-payment 55$ per year

To decide whether to take the medication or not based on a single value of risk reduction (3-year NNT)

Leaman, 2001, UK47

216 participants (41%); 40.5% between 50-69 years; 55% women; mailed questionnaire

Outpatients (single GP); the mean age at which respondents finished education was 16.7 years; 13% on antihypertensive medication; 2% previous MI

Hypothetical; iterative process; anti-hypertensive daily; heart attack; some side effects; some payment

To choose, from 6 different estimates (5-year NNT), the number of lives should be saved to justify lifelong medication; negative framing

Marshall, 2006, UK48

203 participants; median age 65; 13% women; face-to-face interview

Outpatients (13 Practices, No CVD history); mean years of education 15.5; 34.5% on long-term oral medications; 50% had less than 7.5% 5-year coronary risk

Hypothetical; multiple choice experiment; daily; heart disease; rare side effects;

To decide whether to take the medication or not based on a single value of risk reduction (5-year ARR); each one decide 6 times

McAlister, 2000, Canada32

74 participants (51%); mean age 49; 53% women; face-to-face interview

Outpatients (5 GPs and 4 internists, mild hypertension; no overt CVD); 69% educated to more than 12 years; 65% on antihypertensive

Hypothetical; multiple choice experiment; multiple scenarios; antihypertensive daily; death, MI, stroke;

Iterative process to achieve the smallest benefit (ARR) that justify lifelong medication (MCID)

Misselbrook, 2001, UK31

309 participants (102 hypertensive, 207 normotensive) (89%); aged 35-65 years; mailed questionnaire

Outpatients (single practice); patients excluded if illiterate, CVD, diabetes, disability, or mental illness

Hypothetical; single choice experiment; anti-hypertensive-like daily; stroke;

To decide whether to take the medication or not based on a single value of risk reduction (ARR, NNT)

Nexoe, 2005, Denmark37

2326 participants (50%); mean age 47 years; 56% women; face-to-face interview

Community; 26% educated to elementary-level

Hypothetical; single choice experiment; multiple scenarios; daily; death or heart attack; out of pocket payment

To decide whether to take the medication or not based on a single value of risk reduction (NNT)

Nicholson, 1999, UK49

384 participants (53%); mailed questionnaire

Outpatients (1 GP) Hypothetical; single choice experiment; statin-like daily; MI

To choose, from 4 different estimates (5-year NNT), the number of lives should be

25

saved to justify lifelong medication

Sorensen, 2008, Denmark 50

1519 participants (49%); mean age 59; 53.9% women; face-to-face interview

Community; 6% previous heart attack; 17% hypercholesterolemia

Hypothetical; single choice experiment; daily; death or heart attack; mild and harmless side effects; 45€ per year

To decide whether to take the medication or not based on a single value of risk reduction (3-year RRR with baseline risk)

Steel, 2000, UK38

58 participants (58%); 45% 41-65 years old; 58% women

Community (39 practices) Single choice experiment; anti-hypertensive medication

To choose, from a five different estimates (5-year NNT), the number of lives should be saved to justify lifelong medication

Stovring, 2008, Denmark28

1169 participants (37%); aged 40-59; 57% women; face-to-face interview

Community; median years of education 13 years; 37% CVD history

Hypothetical; single choice experiment; multiple scenarios; statin-like daily; fatal heart disease; out of pocket

To decide whether to take the medication or not based on a single value of risk reduction (10-year ARR, NNT, POL)

Trewby, 2002, US33

307 participants (97%); mean age 61 years; 42% women; face-to-face (or telephone) interview

Both (3 groups: 1discharged from CCU; 2no MI history but on medication, 3neither history nor medication)

Hypothetical; single choice experiment; statin-like daily; heart attack; safe

Iterative process to achieve the minimum threshold (POL) that justify lifelong medication; extracted from Figure3