2017
A SINGLE-TECHNOLOGY ASSESSMENT
FreeStyle Libre Flash Glucose Self-Monitoring System
REPORT
1 Executive summary
Title FreeStyleLibreFlashGlucoseSelf‐MonitoringSystem:ASingle‐TechnologyAssessment
Norwegiantitle FreeStyleLibresystemetforegenmålingavblodsukker:enhurtigmetodevurdering
Institution NorwegianInstituteofPublicHealth,DivisionforHealthServices(Folkehelseinstituttet)
CamillaStoltenberg,DirectorAuthors Bidonde,Julia,Researcher(projectleader),
Fagerlund,BeateCharlotte,HealtheconomistFrønsdal,KatrineB,SeniorresearcherLund,UlrikkeHøjslev,Healtheconomist
Robberstad,Bjarne,Seniorhealtheconomist
ISBN 978‐82‐8082‐852‐1Typeofreport Single‐technologyassessment(Hurtigmetodevurdering)No.ofpages 88(includingappendices107)
Client TheCommissionerForumRHA(BestillerforumRHF)Subjectheading(MeSH) BloodGlucoseSelf‐Monitoring,DiabetesMellitus
Citation BidondeJ,FagerlundBC,FrønsdalKB,LundUH,RobberstadB.FreeStyleLibreflashglucoseself‐monitoringsystem:asingle‐technologyassessment.NorwegianInstituteofPublicHealth(Folkehelseinstitutttet).Oslo:NorwegianInstituteofPublicHealth,2017.NorwegianInstituteofPublicHealth(NIPH)Oslo,August2017
2 Executive summary
Executivesummary
Background
Diabetesmellitus(DM)hasbecomeoneofthemostcommonpublichealthproblems
world‐wide.Accordingtothe2014NorwegianPublicHealthreport,diabetesaffectsan
estimated4.3%oftheNorwegianpopulation.Diabetesisametabolicdisorderresulting
fromadefectininsulinproduction,secretion,action,orall.Type1and2arethetwo
maintypes,withtheprevalenceoftype2accountingforthemajority(>85%)ofdiabe‐
tes.ThisassessmentwillfocusonFreeStyleLibre,flashglucosemonitorforinsulin
treatedindividualswithtype1and2diabetes(“Type1and2DM”).
Toachieveproperqualityoflifeandreducelong‐termproblems,peopleareincreas‐
inglyencouragedtotakeanactiveroleinthemanagementoftheircondition.Adequate
treatmentmanagement,aimedattightcontrolofbloodglucose,reducestheriskofthe
long‐termcomplicationsofdiabetessuchasretinopathy,nephropathy,neuropathy,
coronaryheartdisease,ischaemicstrokeandperipheralvasculardisease.‘Manage‐
ment’ofthediseaseshouldbeunderstoodasapackageincludingtestingofbloodglu‐
cose,takinginsulin(i.e.,multipledailyinsulininjections,usinganinsulinpump),using
antihyperglycemicdrugs,oradoptinglifestyleinterventionssuchasdietandphysical
activity.
Inrecentyears,andavailableinEuropesince2014,theFreeStyleLibreSystem‐a
‘wireless’methodusingasensorformonitoringinterstitialfluidglucose‐wasintro‐
ducedtohelpindividualswithtype1and2DMachievebetterglucosecontrol.Thesys‐
tem,unlikeothers,doesnotrequirefingerprickcalibration,sincethatfunctionalityis
embeddedintothecoretechnology.Also,unlikeothersystems,theindividualhasto
takeactiveactiontogetaccesstotherealtimeglucosevalue,byleadingthereceiver
overthesensor.Similarlytoothercontinuousglucosemonitoringoptions,itrelieson
theindividualtotakeactionontheinformationretrieved.
Objective
3 Executive summary
Ourgoalwastoassesstheclinicaleffectiveness,costeffectivenessandsafetyofFree‐
StyleLibreforindividualswithtype1and2DM.
Methods
Weconductedasystematicreviewaccordingtostandardmethodstosummarisethe
evidence.ThestudypopulationswereinsulintreatedindividualswithType1or2DM,
theinterventionwasFreeStyleLibre,andtheoutcomeswereHbA1c,hypoandhyper‐
glycaemia,qualityoflife,patientsatisfaction,pain,andadverseevents.
Wesearcheddatabases,trialregistries,healthtechnologyassessmentagenciesweb‐
sitesandgreyliteraturefrominceptiontoJanuary2017withnolanguagerestrictions.
Tworeviewersindependentlyscreenedthetitlesandabstractsofallrecordsidentified
bysearches,discussedanydiscrepanciesandsolvedthembyconsensus.Weobtained
fulltextcopiesofallstudiesdeemedpotentiallyrelevantandthesametworeviewers
independentlyassessedtheseforinclusion;solvinganydisagreementsbyconsensus.
Onereviewerextracteddatarelatingtostudydetails,participants,intervention,and
comparator,usingapiloted,standarddataextractionform.Asecondreviewerchecked
dataextractionandanydisagreementsweresolvedbyconsensus.Theassessmentof
themethodologicalqualityofeachincludedstudywasbasedontheCochraneCollabo‐
rationriskofbiastool.Qualityassessmentofevidencewascarriedoutindependently
bytworeviewers.Wesolvedanydisagreementsbyconsensus.Meta‐analysiswascon‐
sideredasuitableanalysisforthedataidentified,despiteheterogeneity.Forsomeout‐
comesweemployedanarrativesynthesis.
Assessmentofcosteffectiveness
Weassessedthecost‐effectivenessestimatesprovidedbythesubmitterofFreeStyleLi‐
brecomparedtoself‐monitoringbloodglucose(SMBG)forindividualswithtype1and
2DM.Thesubmitterusedacommerciallyavailablecost‐effectivenessmodel,IMSCORE
diabetesmodel(IMSCDM)forthisassessment.Themodelisinternetbased,witha
Markovapplication,forindividuals>18years.Theinteractivesimulationpredictsthe
long‐termhealthoutcomesandcostsassociatedwiththemanagementoftype1and2
DM.Themodelconsistof17sub‐modelsdesignedtosimulatediabetesrelatedcompli‐
cations,nonspecificmortality,andcostsovertime.Asthemodelsimulatesindividual
patientsovertime,itupdatesriskfactorsandcomplicationstoaccountfordiseasepro‐
gression.However,thismodelreceivedfromthesubmitter,lackstransparency,and
madeitdifficulttogainafirmunderstandingofthefactorsthatdeterminehowpatients
4 Executive summary
progressthroughthemodel,assumptionsandparameterseffectoutcomesandtoas‐
sessthevalidityofthemodel.BecausetheNorwegianInstituteofPublicHealthdidnot
havecompleteaccesstothemodel,itwasnotpossibletoperformafullassessmentof
themodelortomodifyunderlyingassumptionsandparametersinordertoinde‐
pendentlyassesstheimpactonreportedresults.Furthermore,thedocumentation
packagedidnotincludeanysensitivityanalysis,whichisessentialforconsideringthe
validityandrobustnessofresultsfromeconomicevaluations.
Results
Weincludedtworandomizedcontrolledtrials(RCTs)inthereview.Thesestudiescom‐
paredFreeStyleLibretoSMBG.Also,wefoundseveralpublicationsinvestigatingthe
accuracyofthedevice,however,thestudydesignsofthesestudies(singlearm)didnot
meettheinclusioncriteriaofthisevaluationand,althoughwecompiledthemforinfor‐
mation,theywereexcludedfromthesynthesis.Theinformationderivedfromthese
singlearmstudiesarepotentiallyimportanttovalidatethesensitivityandspecificity
estimatesofFreeStyleLibre.Inaddition,wefoundotherEuropeanassessmentscon‐
ductedinthepast6to8months.TheincludedRCTsreporteddataonmiddleaged
adultsfromEuropeancountrieswithtype1and2DMat6monthspostintervention.
Weratedthestudies’riskofbiasasuncleartohighrisk.
MainfindingsfromthesetrialsarethatFreeStyleLibremayslightlyimprovetreatment
satisfaction,timespentwithglucoseinrange3.9to10mmol/L,numberofnocturnal
eventswithglucoselevels<3.1mmol/Lwithin7h,andtimespentwithglucoselevels
>13.0mmol/LincomparisontoSMBG.FreeStyleLibreleadtolittleornodifferencein
qualityoflifeandHbA1clevelincomparisontoSMBG.Theevidenceisuncertainabout
whetherFreeStyleLibreleadstoanimprovementintimeandeventswithglucose
<3.9mmol/Lwithin24h,timewithglucose<3.1mmol/Latnightwithin7hours,and
timewithglucose>10mmol/L.
Thesubmittedeconomicmodelrunsa40‐yeartimehorizon.Thesubmitter´sbasecase
suggestedthatthetechnologyisdominantforindividualswithtype1DM,i.e.thatFree‐
StyleLibreisacheaperandmoreeffectivetechnology.Accordingtosubmitter´sbase
case,individualswithtype2DMtheincrementalcost‐effectivenessratio(ICER)was
calculatedtobeNOK235,673perQALY(wholestudypopulation)andNOK243,434
perQALY(under65years).Asthemodelreceivedbythesubmitterwasneithersuffi‐
cientlytransparentnorsufficientlyflexibletoallowchanges,wehavenotbeenableto
producealternativeincrementalcost‐effectivenessratios(ICERs).Fromahealthcare
5 Executive summary
perspective,thesubmitterhascalculatedabudgetimpactfortype1DMtohaveatotal
addedcostthefifthyearafteradoptionofthetechnology.Further,thesubmittercalcu‐
latedabudgetimpactfortype1and2DMthatleadtoacostsavingonthefifthyearaf‐
teradoptionofthetechonology.Thesubmitterdidnotcalculateabudgetimpactfor
type2DMonly.
Weestimatedthat,fromahealthcareperspective,theannualcostsfiveyearsafterin‐
troductionwouldbeNOK186millionaddedcostandNOK91,7millionsavedcostfor
type1and2DMalone,respectively,andNOK94millionaddedcostfortype1and2
DMcombined.
Conclusions
Overall,theevidencefortheinterventionofinterestwaslimitedbutsuggeststhatFree‐
StyleLibreincreasestreatmentsatisfaction,reducessomehypo‐andhyperglycaemic
measures(increasestimewithglucoseinrange3.9to10mmol/L,reducestimeand
numberofeventswithglucose<3.9in24hours,numberofglucose<3.1nightevents
andtimewithglucose>13mmol/L)andhassimilarseriousadverseeventsthanSMBG,
withoutdifferencesinotheroutcomesincludingHbA1candqualityoflife.
Thequalityoftheincludedstudieswasgenerallylowandtherewereonlytwosmall
studiesincludingmiddleagedadults.
Severalinconsistenciesleadustoquestiontheresultofthesubmittedhealtheconomic
report.Specifically,thesubmittedmodelincludedseveralinputdatathatdidnotmatch
theinputdatadescribedinthesubmitteddocumentationpackage,andnordiditmatch
theinputdatafoundinotherliterature.
Themostchallengingissueisthatthemodelisnotsufficientlytransparentorflexible,
sincewedidnothaveaccesstothecompletemodel.Therefore,wewerenotabletoas‐
sesshowthepossibleadjustmentswouldaffecttheresultsprovidedbythesubmitters.
Suggestedresearchpriorities
IndependentresearchforFreeStyleLibrewillbeimportant
Diabetesaffectsthelifeofchildren,adolescentsandtheircaregiversinmany
ways,aswellaspregnantwomen.Independentresearchincludingthesegroups
iswarranted
6 Executive summary
TheclinicaleffectivenessofFreeStyleLibreneedstobeinvestigatedindifferent
conditions,forexample,amongindividualswithpoorself‐monitoring
adherence,newlydiagnosed,impairedawarenessofhypoglycaemia,andin
additiontotrainingandeducationcomponents
FreeStyleLibrecomparedtoothercontinuousmonitoringsystemsiswarranted
Painisamajordeterminantofdiabetestreatmentadherence,especiallyfor
children,anditshouldbeincludedasanindividualoutcomeinfuturetrials
Futuretrialsshouldincludelongertermfollowupandqualityoflifeoutcome
assessmentsatvariouspointstoinformimprovedclinicalandcost
effectivenessmodelling
7 Sammendrag (norsk)
Sammendrag(norsk)
Bakgrunn
På verdensbasis er diabetes mellitus (DM) i dag blant de mest vanlige helseproble-
mene. I følge Folkehelserapporten fra 2014 regner man med at 4.3%avNorgesbe‐
folkningenerrammet.DMerenmetabolsksykdomforårsaketentenavmanglendein‐
sulinproduksjonellermanglendeinsulinresponsellerenkombinasjonavdisse.DM
type2stårforflerpartenavtilfellene(>85%).Idennerapportenharvivurdertbrukav
FreeStyleLibre‐systemetformålingavblodsukker(glukose)hospersonermedtype1
og2DMsombehandlesmedinsulin.
Foråoppnåbedrelivskvalitetogforhindrekomplikasjonerovertiderdetenfordelat
pasiententarenmestmuligaktivrolleiåovervåkeogbehandleegensykdom.God
oppfølgingogkontrollavglukosenivåreduserernemligrisikoenpålangsiktforkom‐
plikasjonergrunnetdiabetes,somforeksempelretinopati,nefropati,nevropati,(syk‐
dom/skadeinetthinne,nyreogperiferenerver),samthjerte‐ogkarsykdommerog
hjerneslag.Behandlingenavdiabetesbestårien«pakke»avuliketiltak,sominnebærer
testingavglukosenivåiblod,inntakavinsulin(foreksempelvedåinjisereinsulinflere
gangerperdagellervedbrukavinsulinpumpe),brukavblodsukkersenkendemedika‐
menterogendringilivsstilmedhensyntildiettogfysiskaktivitet.
Idesisteparårene,ogsiden2014iEuropa,harFreeStyleLibre‐systemetværttilgjeng‐
eligpåmarkedet.Dettenyesystemeterenmetodesombenytterensensorforåmåle
glukosenivåetideninterstitiellevæsken,somlesesavtrådløst.Hensikteneråoppnå
bedrekontrollavblodsukkernivåethospersonermedtype1og2DM.Tilforskjellfra
andremålemetoderkreverikkedettesystemetfingerstikkforkalibrering,dadenne
funksjonenerintegrertiselveteknologien.Itillegg,ogsåuliktandremetoder,måbru‐
kerenselvføreavleserennærsensorenogskannedenneforåfåviteaktueltglukose‐
nivå.Metodeneraltsåavhengigavatbrukerenbidrarselv,ogisåmåteerdettesyste‐
mettilsvarendeandremetoderforkontinuerligglukosemonitorering.
8 Sammendrag (norsk)
Mål
Måletmeddennerapportenvaråvurderekliniskeffektogsikkerhet,samtkostnadsef‐
fektivitetavFreeStyleLibrehospersonermedtype1og2DM.
Metode
Viutførteensystematiskoversiktihenholdtilstandardmetodikkforåoppsummere
kunnskapsgrunnlaget.Viinkludertestudierhvorpopulasjonenehaddetype1ellertype
2DMoghvorintervensjoneninnebarbrukavFreeStyleLibre.Depredefinerteutfalls‐
målenevarHbA1c,hypo‐oghyperglykemi,livskvalitet,pasienttilfredshet,smerteog
uønskedehendelser.
Visøkteidatabaserogstudieregistere,menogsåpånettsidenetilandreorganisasjoner
somgjørmetodevurderinger,samtettergrålitteraturpublisertfremtiljanuar2017
utenbegrensningmedhensyntilspråk.Toforskeregikkuavhengigavhverandregjen‐
nomtitlerogabstrakterpåalletreffeneidentifisertvialitteratursøket.Uoverensstem‐
melserblediskutertinntilkonsensusvaroppnådd.Viinnhentetmuligrelevantepubli‐
kasjonerifulltekst.Etterattoforskerehaddelestgjennomdisseuavhengigavhver‐
andrebestemteviomviskulleinkluderestudienimetodevurderingen.Ogsåpådette
trinnetløsteviuenighetergjennomdiskusjon.Mensénmedarbeiderekstraherteinfor‐
masjonomstudiene,detvilsiomdeltakerne,intervensjonen,komparatoren,utfallog
effektestimatervedbrukavetvelutprøvdstandardekstraksjonsskjema,sjekketenan‐
nenmedarbeiderdeekstrahertedataene.Viløsteuoverensstemmelsersombeskrevet
fordetidligeretrinneneiprosessen.Vurderingavdenmetodologiskekvalitetenpå
hveravdeinkludertestudienebleutførtvedbrukverktøyet for å vurdere risiko for
systematiske skjevheter utvikletavCochraneCollaboration(RiskofBias).Bedømmel‐
senavkvalitetenavkunnskapsgrunnlagetblegjortavtoforskerehverforseg.Ogsåher
varuenighetløstsombeskrevettidligereiavsnittet.Vivurderteatdataeneegnetsegtil
åinngåimetaanalyser,tiltrossforheterogenitetmellomeffektestimatene.Fornoenut‐
fallsmålvarsyntesengjortnarrativt.
Evalueringavkostnadseffektivitet
VievaluertekostnadseffektivitetentilFreeStyleLibresammenlignetmedselvkontrol‐
lertblodsukkermåling(SMBG)forpersonermedtype1og2DM.Abbottbrukteen
kommersielttilgjengeligkostnadseffektivitetsmodell,IMSCOREdiabetesmodellen(IMS
CDM)forevalueringen.ModellenerinternettbasertogMarkov‐basert,forpersoner
over18år.Deninteraktivesimuleringenpredikererdelangsiktigehelsemessigeutfal‐
leneogkostnadeneknyttettiladministreringenavtype1og2DM.Modellenbestårav
9 Sammendrag (norsk)
17delmodellerdesignetforåsimulerediabetesrelatertekomplikasjoner,uspesifisert
dødelighetogkostnaderovertid.Sidenmodellensimulererindividuellepasienterover
tid,oppdatererdenrisikofaktorerogkomplikasjonerforåtahensyntilsykdomspro‐
gresjon.ModellenvimottokfraAbbotterikketransparent,ogderforvardetvanskelig
åfåensolidforståelseavfaktorenesombestemmerenkohortsfremganggjennommo‐
dellutviklingen,antagelseneogparametereffektresultateneogforåvurderemodellens
validitet.Folkehelseinstituttethaddeikketilgangtildenfullstendigemodellen,ogdet
varderforumuligåutføreenfullstendigevalueringavmodellenelleråendreunderlig‐
gendeforutsetningerogparametereforåkunnevurdereeffektenavrapporterteresul‐
tater.Videreinkluderteikkedokumentasjonspakkennoensensitivitetsanalyse.
Resultater
Idennehurtigmetodevurderingeninkludertevitorandomisertekontrollertestudier
(RCTer).DissestudieneharsammenlignetFreeStyleLibremedSMBG.Itilleggfantvi
flerepublikasjonersomomhandletutstyretsmålenøyaktighet,mengrunnetstudiede‐
signet(single‐arm)ograpportensformålbleikkedisseinkludert.Dogerkjennervibe‐
tydningenavdissesingle‐armstudienemedtankepåatdevisersensitivitetenogspesi‐
fisitetenavFreeStyleLibresommetodeforåmåleglukosekonsentrasjon.Deinkluderte
RCTenerapportertefunnpåmiddelaldrendevoksnefraulikeeuropeiskelandmedtype
1og2DMseksmånederetteratintervensjonenpåbegynte.Ettervårvurderinghadde
studienefra«uklar»til«høy»risikoforsystematiskeskjevheter(riskofbiases).
HovedfunnenefradeinkludertestudieneeratFreeStyleLibremuligensforbedrerbru‐
kerstilfredshet,tidtilbraktmedglukosenivåmellom3,9og10mmol/l,antallnattlige
hendelsermedglukosenivåer<3.1mmol/liløpetav7timerogtidmednivåer>13.0
mmol/lsammenlignetmedSMBG.ImidlertidmedførerFreeStyleLibrelitentilingen
forskjellmedhensyntillivskvalitetognivåavHbA1csammenlignetmedSMBG.Vier
usikrepåomFreeStyleLibreførertilforbedringavtidoghendelsermedglukosenivå
<3.9mmol/Liløpetavettdøgn,tidmednivåer<3.1mmol/Lmåltpånatten(7timer),
samttidtilbraktmedglukosenivå>10mmol/l.
Denøkonomiskemodellen,innsendtavAbbott,haddeen40årstidshorisont.Deninn‐
sendtemodellentyderpåatFreeStyleLibreerdominantforindividermedtype1DM,
detvilsikostnadseffektiv.IfølgeAbbott,forpersonermedtype2DMbledeninkre‐
mentellekostnadseffektivitetsratioen(ICER)beregnettil235673kronerperQALY
(helestudiepopulasjonen)og243434kronerperQALY(under65år).Sidenmodellen
10 Sammendrag (norsk)
levertavAbbottikkevartransparent,harviikkehattmulighettilåproduserealterna‐
tiveinkrementellekostnadseffektivitetsratioer(ICER).Medutgangspunktiethelsetje‐
nesteperspektiv,estimerteAbbottbudsjettkonsekvensfortype1DMtilågimerkostna‐
derfemteåretterteknologienblirinnført.VidereestimerteAbbottbudsjettkonsekvens
fortype1og2DMtilåværekostnadsbesparendefemteåretterteknologienblirinn‐
ført.Abbottestimerteikkebudsjettkonsekvensfortype2DMseparat.Fraethelsetje‐
nesteperspektivestimerteviimidlertidatdentotaleårligekostnadenfemåretterin‐
troduksjonavFreeStyleLibrehenholdsvisvilgi186millionerkronermerkostnaderfor
type1DMaleneog91,7millionerkronerspartekostnaderfortype2DMalene.Ved
kombinasjonavtype1ogtype2DMvilintroduksjonenavFreeStyleLibregiomkring
94millionerkronermerkostnaderfemåretterintroduksjonen.
Konklusjoner
Genereltsettvarkunnskapsgrunnlagetbegrenset,menresultatenekantydepåat
FreeStyleLibreøkerpasienttilfredsstillelseogatdetreduserernoevarighetoghyppig‐
hetavhypo‐oghyperglykemi‐episoder.FreeStyleLibreøkermuligenstidtilbragtmed
glukosenivåmellom3.9og10mmol/l,reduserertidogantallhendelsermedglukose‐
nivå<3.9mmol/loveretdøgn,hendelsermedglukosenivo<3.1mmol/l‐iløpetavnat‐
tenogtidtilbagtmednivåerglukose>13mmol/l.FreeStyleLibremedførertilsvarende
antalluønskedehendelsersomSMBG.Detersmåelleringenforskjellermedhensyntil
deandreutfallene,somHbA1coglivskvalitet.Kvalitetenpådetoinkludertestudiene
vargenereltsettlavoginkludertevoksnemiddelaldrendedeltakere.
Deterflereuoverensstemmelsermellomforutsetningenenevntidenskriftligedoku‐
mentasjonspakken,iannenliteraturogantakelsenesomfaktiskerbruktidenhelse‐
økonomiskemodellen.Itilleggerdeninnsendtehelseøkonomiskemodellenikketrans‐
parent.Vihaddeikkefullstendigtilgangtilhelemodellen,ogdetførtetilstorusikker‐
hetomkringmodellensstrukturogkvalitetenavinnholdet.Pågrunnavdetteerdet
ikkemuligåsehvordanvårejusteringerogkorreksjonerhaddepåvirketresultatet.
Forslagtilvidereforskning
DeterviktigatdetgjøresuavhengigforkningpåFreeStyleLibre
Diabetesrammerogsåbarnogungdomogdesomharomsorgfordem,og
gravidekvinner.Deterderforbehovforuavhengigforskning,sominkluderer
dissegruppene
EffektavFreeStyleLibremåundersøkesunderulikeaktuelleforhold,somfor
eksempelblantpersonersomikkefølgeroppsinsykdomellersomhar
11 Sammendrag (norsk)
vanskelighetermedutførelsenavsinblodsukkerkontroll,hosdemsomnyliger
blittdiagnostisertmeddiabetesoghospersonersomikkeerklaroverellerikke
følereventuellhypoglykemi.Forholdknyttettiltreningogopplæringerdet
ogsåviktigåfåmerkunnskapom.
FreeStyleLibremåsammenlignesmedandremetoderforkontinuerlig
blodsukkermonitoring
Smerteervesentligmedtankepåtilslutningtilethvertbehandlingsprogramfor
diabetes,spesielthosbarn,ogdettebørinkluderessometseparatutfallsmåli
fremtidigestudier
Viderebørstudienefølgeoppdeltakerneoverlengretid,samtmåle
livskvalitetsutfallveduliketidspunktforåbidratilbedremodelleringavklinisk
effektogkostnadseffektivitet.
12 Table of contents
Tableofcontents
EXECUTIVESUMMARY 2
SAMMENDRAG(NORSK) 7
TABLEOFCONTENTS 12
ABBREVIATIONS 14
PREFACE 15 Whatisasingle‐technologyassessment? 15 Objective 15 Log 15 History 16 Projectteam,consultantsandstakeholders 16
BACKGROUND 18 Thetechnology 18 Regulatorystatus(CE‐marking)andmarketaccessofthetechnology 18 Descriptionanduseofthetechnology 18 Descriptionofthehealthcondition 21
CLINICALEFFECTIVENESSANDSAFETY 25 Issueaddressed 25 Objective 25 Methods 25 Results 32 Characteristicsofincludedtrials 34 Riskofbiasinincludedstudies 35 EffectofFreeStyleLibre 38
COST‐EFFECTIVENESSANALYSIS 51 Methods for evaluating submitted cost-effectiveness models 51 Publishedcost‐effectivenessevaluationsidentifiedbythesubmitter 52 Populationandcomparatorinthesubmittedreport 54 Typeofanalysisanddecisionmodelofthesubmittedreport 54 Generalcommentsonthesubmittedhealtheconomicanalysis 56 Clinicalandepidemiologicaldata 57 Efficacy 58
13 Table of contents
Costs 59 Healthrelatedqualityoflife(HRQL) 62 NIPHcommentsonsubmitterparametersandinputdata 63 Cost‐effectivenessresults 65 Budgetimpactanalysis 66
DISCUSSION 71 Clinicaleffectivenessandsafety:summaryofmainresults 71 Overall completeness and applicability of evidence 72 Quality of the evidence 73 Other European assessments 73 Implications for clinical practice 76 Cost-effectiveness 84 KeyConclusions 87
APPENDICES 89 References 89 Appendix1.Glossaryofterms 95 Appendix2.Searchstrategy 97 Appendix3.Completed(publishedandunpublished)singlearmstudies 101 Appendix4.FreeStyleLibreregistryrecords 102 Appendix5.Includedstudiesinclusionandexclusioncriteria 104 Appendix 6. Results from Meta-Analysis 105
14 Abbreviations
Abbreviations
Technicaltermsandabbreviationsareusedthroughoutthisreport.Themeaningisusuallyclearfromthecontext,butaglossary(seeAppendix1)isprovidedforthenon‐specialistreader.CI ConfidenceintervalCPI ConsumerpriceindexDM DiabetesmellitusDRGs Norwegiandiagnoses‐relatedgroupsGP GeneralpractitionerHELFO TheNorwegianHealthEconomicsAdministration.HRQL HealthrelatedqualityoflifeHTA HealthtechnologyassessmentICER IncrementalcosteffectivenessratioIMSCDM IMSCoreDiabetesModelMD MeandifferenceMmol MilimolesMol MolesSMBG SelfmonitoredbloodglucoseNIPH NorwegianInstituteofPublicHealthNOK NorwegianKronerQALY QualityadjustedlifeyearRCT RandomisedcontrolledtrialRR RiskratioSD Standarddeviation
15 Preface
Preface
Whatisasingle‐technologyassessment?
Asingle‐technologyassessmentisoneofaseriesofhealthtechnologyassessment(HTA)productsthatcanbemandatedin“TheNationalSystemforIntroductionofNewHealthTechnologies”withintheSpecialistHealthServiceinNorway(https://nyem‐etoder.no/).Withinthissystem,theCommissionerForumRHA(“BestillerforumRHF”),wherethefourRegionalHealthAuthoritiesarerepresented,evaluatessubmittedsuggestionsanddecidesonwhichtechnologiesshouldbeassessedandthetypeofassessmentneeded.Inasingle‐technologyassessment,thetechnology(apharmaceuticaloradevice)isas‐sessedbasedondocumentationsubmittedbythecompanyowningthetechnology,ortheirrepresentatives(“thesubmitter”).TheHTAunitoftheNorwegianInstituteofPublicHealth(NIPH)receivesandevaluatesthesubmitteddocumentation,butisnotthedecision‐makingauthority.Single‐technol‐ogyassessmentsconductedatNIPHarepublishedonourwebsite(www.fhi.no)andonhttps://nyemetoder.no/
Objective
Toassesstheclinicaleffectiveness,cost‐effectivenessandsafetyofFreeStyleLibre(in‐terstitialmeasurementofglucose)forinsulintreatedindividualswithdiabetestype1and2.
Log
WereceivedtheFreeStyleLibrecomission,ID2016_044,onDecember23,2016.TheCommissionForumrequestedtheNIPHHTAUnittoperformaclinicaleffectivenessandsafetyassessmentalongwithacost‐effectivenessanalysisofthissingle‐technologyforthemanagementofinsulindependentindividualswithdiabetestype1and2.Infor‐mationonthecommissioncanbeseenherehttps://nyemetoder.no/metoder/system‐freestyle‐libre‐for‐egenmaling‐av‐blodsukker
16 Preface
Date Correspondence
September 23, 2016 Publication of horizon scanning report on this device
October 24, 2016 The commissioning forum commissioned a single technology assessment
Oct. 2016 – Dec 2016 Dialogue and meeting with technology manufacturer
December 23, 2016 Valid submission acknowledged
April 26 – 27, 2017 Clinical experts, and stakeholders draft reviewing
May 12, 2017 Norwegian Institute of Public Health external review process
May 25, 2017 Norwegian Institute of Public Health internal review process
June 2, 2017 End of 180 days evaluation period –Report Submitted
June 8, 2017 Report available at FHI website
July-Aug 2017 Amendments to report considered (i.e. health economic)
August 21th 2017 Amended report available at FHI website
History
Reportfirstsubmissiontocommissionerforum:June2nd,2017Amendedversion:August21th2017:themethodsusedinoriginalreportremainintact.Amendmentsweredoneasfollow:removalofinformationdeemedconfidentialbythesubmitter,furtherexplanationoftransparencyissuesencounteredwiththeIMSCoremodel,andelaborationofdiscrepanciesbetweenthecost‐effectivenessanalysisandbudgetimpactanalysis.Inclusionofin‐textevidencederivedfromsinglearmstudyde‐signs(previouslyinappendices),andupdatedinformationonEuropeancountriesHTAs.
Projectteam,consultantsandstakeholders
Norwegian Institute of Public Health team
Julia Bidonde (project leader), Researcher
Beate C. Fagerlund, Health Economist
Katrine B. Frønsdal, Senior Researcher
Ulrikke Højslev Lund, Health Economist
Bjarne Robberstad, Senior Health Economics
Ingrid Harboe, Information Specialist
Ingvil Sæterdal, Acting Head of Unit
Acknowledgements
Torunn Tjelle, Researcher
Elisabet Hafstad, Information Specialist
Doris Tove Kristoffersen, Statistician
Arna Desser, Senior Health Economist Nils Vӧlker, Health Economist fellow
Angieszka Anna Zachariassen, Legal services
17 Preface
In process consultants Åsne Bakke, Chief Physician, Stavanger University Hospital
Kristin Aakre Moberg, Department Chief Physician, Haukeland University Hospital
Torstein Baade Rø, Head of Department, St. Olavs Hospital/Associate Professor, NTNU
Henrik Underthun Irgens, Chief Physician, Haukeland University Hospital
External consultants Tore Julsrud Berg, Chief Physician, OUS/ Associate Professor, UiO.
Member of board of medical advisors for the Norwegian Diabetes Association
Sven Magnus Carlsen, Professor, NTNU.
Stakeholders who provided feedback on this assesment Norwegian Diabetes Association
Ragnhild and Andreas Jansson
SigneAgnesFlottorpDepartmentdirector
IngvilSæterdalActingHeadofUnit
JuliaBidondeProjectleader
18 Background
Background
Thetechnology
Nameofthetechnology:FreeStyleLibre,FlashGlucoseSensorMonitorSystem(“Free‐StyleLibre”)Manufacturerwhichsubmittedtheapplicationandprovidedthedocumentationpack‐age:AbbottNorgeAS(Leverandør),Postboks1,1360Fornebu,Norge(“submitter”).
Regulatorystatus(CE‐marking)andmarketaccessofthetechnology
FreeStyleLibrehasEuropeanConformity(ConformitéEuropéenne)orCE‐markingClassIIb,notificationbytheBritishStandardsInstitution(No597686),UnitedKing‐dom.ClarificationofCEMarkingofMedicalDevicescanbefoundathttp://www.ce‐marking.com/medical‐devices‐class‐IIb.htmlFreeStyleLibreisavailableinNorway,butisnotreimbursedbytheNorwegianhealthcaresystem(SpecialistHealthCare)orthroughtheNationalSocial(Security)Insurance(Folketrygden).
Descriptionanduseofthetechnology
FreeStyleLibre(informationbelowtakenfromthesubmitter’spackageandweb‐
site)
FreeStyleLibreflashglucosemonitorisasystemthatreliesonasubcutaneousglucosesensor,usuallyplacedontheupperarm,thatmeasuresandcontinuouslystoresglucosereadingsdayandnight(seeFigures1and2below).Glucoselevelsarecheckedby‘scanning’thesensorwithareaderobviatingtheneedforregularself‐monitorbloodglucose(SMBG)testing.Itreportscurrentglucoseconcentration,glucosetrendanddisplaystheprevious8hoursasatrend;FreeStyleLibreupdatesreadingseveryminuteandstoresdataevery15minutes.Ahand‐heldreaderisusedtoscanandretrieveinformationfromthesensor.Thereadercancapturedatafromthesensorwhenitiswithin1to4cmdistanceofthesensor.Variousreportsareavailablefromthereader: Logbook:individualglucosereadingsanduser‐enterednotes
19 Background
Dailygraph:dailyoverviewofglucosereadings,includinghowtheyfallwithinthetargetglucoserange
Averageglucose:averageglucosereadingsalongwithfour6‐hoursperiodsduringtheday
Dailypatterns:indicateswhenglucoselevelsareinthetargetrangeandthevariabilityofglucoselevels
Timeintarget:indicatesthepercentageoftimeglucosereadingsareinthetargetrangeandaboveorbelowthetargetrange
Lowglucoseevents:indicatesthenumberoflowglucoseeventsatfourdifferenttimesoftheday
Sensorusage:indicatesaveragenumberofscansperdayandwhatpercentageofglucosedatahasbeencapturedbythesescans
FreeStyleLibreisfactory‐calibrated,andthesensorscanbewornforupto14days.Thesensoriswaterresistantinupto1meterofwaterforamaximumof30minutes;itcanbewornwhilebathing,showering,swimmingandexercising.
Figure1.FreeStyleflashglucosesystem(imageusedwithsubmitter’spermission)
Figure2.FreeStyleapplicationandscanning(imageusedwithsubmitter’spermission)FreeStyleLibreinitialpackageincludesthefollowingcomponents:AreaderKit
1FreeStyleLibreReader 1USBcable 1Poweradapter 1QuickStartGuide UserManual
20 Background
ASensorKit
DisposableSensor:ithasathin,sterilefilament(0.4mmwide,insertedapproximately5mmunderskin)attachedtoasmalldisk(30mmx5mm)
Medicalgradeadhesiveisusedtokeepthesensorinplaceontopoftheskinonceapplied
1Sensorapplicator 1Alcoholwipe ProductLeaflet
OptionalsofwareSpeciallydesignedsoftware,givespeoplethepossibilitytodownloaddatatoacom‐puterandtocreateaseriesofreportsthatprovideafullglycemicpictureacrossvari‐oustimeframes.Reportsarepresentedingraphicalformatsthatareeasyforpeopletointerpret.Thereportsareasfollow:
Snapshot DailyPatterns GlucosePatternInsights MealtimePatterns MonthlySummary WeeklySummary DailyLog ReaderDetails
ContextforUse(fromthesubmitter’spackage)
FreeStyleLibreisindicatedformeasuringinterstitialfluidglucoselevelsinpeopleaged4andolderwithdiabetesmellitus.Theindicationforchildren(age4‐17)islimitedtothosewhoaresupervisedbyacaregiverwhoisatleast18yearsofage.FreeStyleLibreisdesignedtoreplaceSMBGofdiabeteswiththeexceptionslistedbelow.
Duringtimesofrapidilychangingglucoselevels,interstitialglucoselevels,asmeasuredbythesensorandreportedascurrent,maynotaccuratelyreflectbloodglucoselevels.Whenglucoselevelsarefallingrapidily,glucosereadingsfromthesensormaybehigherthanbloodglucoselevels.Conversely,whenglucoselevelsarerisingrapidly,glucosereadingsfromthesensormaybelowerthanbloodglucoselevels.
Inordertoconfirmhypoglycaemiaorimpendinghypoglycaemiaasreportedbythesensor.
IfsymptomsdonotmatchFreeStyleLibrereading.Donotignoresymptomsthatmaybecausedbylowbloodglucoseorhighbloodglucose.
AdvantagesanddisadvantagesofFreeStyleLibre
21 Background
TheadvantageoftheFreeStyleLibreisthecontinuousprovisionofinformationabout
interstitialglucoseconcentrationthatcanfacilitateadjustinginsulindosage.The
technologyisfactorycalibrated,whichmeanstheindividualdoesnothavetoperform
dailySMBGbyfingerprick,thesensorissmallandcaneasilybehiddenunderclothing,
anditiswaterresistantwhichcanbeseenasanadvantageforthosewhoenjoywater
activities.
DisadvantagesofFreeStyleLibrearepotentialskinirritation,andassociatedcosts
(sensorhastobereplacedevery14days);therecanbesomedelayinthe
measurement,whichmayimpedeoptimalmonitoring.Althoughahighbodymass
indexmaynotinfluencereadings,accordingtoclinicalexpertsopinion,averynarrow
subcutaneousspaceinunderweightindividualsmayperhapsputlimitationstothe
sensoruse.Unlikeotherrealtimemonitors,FreeStyleLibredoesnothaveanalarm,
anditdoesnotworkinsynchronisationwithaninsulinpump.
Descriptionofthehealthcondition
Diabetesmellitus(DM)isametabolicdisorderresultingfromadefectininsulinpro‐
duction,secretion,action,orall.ThenumberofpeoplewithDMishighandrisingin
everycountry‐withthenumberofadultswithDMhavingmorethandoubledover
nearlythreedecades(1).Thelatestestimatesshowaglobalprevalenceof382million
peoplewithDMin2013,anumberthatisexpectedtoriseto592millionby2035(2).
AcrossEuropethereareabout60millionpeoplewithDMandthenumberisexpected
toincreaseto71millionby2040(3).
InNorway(4):
Approximately200,000‐220,000people(4.3%percentoftheNorwegianpopulation)havebeendiagnosedwithdiabetes,
BasedoncalculationsfromtheNorwegianPrescriptionDatabase,inNorway,approximately28,000people(0.6%ofthepopulation)havetype1diabetes,
Approximately28%ofpeoplewithtype2diabetesaretreatedwithdietandexercise,
InternationalstudiesandunpublisheddatafromNorwaysuggestthatmanypeoplearelivingwithundiagnoseddiabetes,
Theprevalenceandincidenceoftype2increaseswithagetoapeakat80years, Norwayisamongthecountrieswithhighernewcasesoftype1DMinchildren
peryear.
22 Background
Theaetiologicalclassificationofdiabeteshasbeenacceptedworldwide;type1and
type2DMarethetwomaintypes,withtheprevalenceoftype2accountingforthema‐
jority(>85%)oftotaldiabetes.Type1isthemostcommonformofdiabetesamong
children,butitcanalsodevelopamongadults.Type1DMrequiresinsulin;peoplewith
type2DMcanbetreatedwithoralmedicationbutmayalsorequireinsulin.Oftenmod‐
ificationsindietandlifestylehelptomanagethediseaseforindividualswithtype2
DM.OthertypesofDMaremonogenic,secondaryandgestational.Monogenicdiabetes
occurswhenthereisachangeinasinglegene.Togethermonogenicdiabetesaccount
for1‐2%ofalldiabetescases,anddependentonthegeneinvolvedthetreatmentisei‐
therinsulinorsulfonylureatablets(5).Secondaryformsofdiabetes(secondarytopan‐
creaticdiseaseoradministrationorcertaindrugs)accountfor1to2%ofalldiabetes
(6).Gestationaldiabetes(5‐10%ofpregnancies)representsariskfactorforfuturede‐
velopmentofDM(7‐9).
Childrenandadolescent
Type1diabetesisthemostcommontypeseeninchildren,andismostcommonlyfirst
diagnosedintheteenageyears.Type2diabetesinchildrenandadolescentsisarela‐
tivelynoveldiseasefacingpaediatrichealthcareproviders(10).Type2isbecomingin‐
creasinglymoreprevalentinyoungerpeople,andmaybemoreinpeopleofSouth‐
Asian,AfricanCaribbeanorMiddleEasterndescent.Childrenwithadiagnosisofdiabe‐
tesoftenpresenttothehealthservicewithissuessuchashypoglycaemia,hyperglycae‐
mia,ordiabeticketoacidosis(11).Diabetesaffectschildren’slifeincludingschoollife,
dailyactivities,theiracademicachievementsandpersonalaspirations.Forexample,
theymaybeaffectedbylackoffulltimeschoolnurses,lackofteacherknowledgeofdi‐
abetes,lackofaccesstodiabetestools,lackoffreedomtoperformdiabetesself‐care,
lackofnutritionalinformationincafeterias,orlackofcommunicationbetweenparents
andschoolpersonnel(12).
Childrenwithtype1DMneedmultipledailymeasurementsoftheirbloodglucose.In
smallchildren,parentsorothercaregiversareresponsibleforthetreatmentandthe
monitoringoftheirdiabeteswhichcaninterferewithdailyactivitiesandwork.Fur‐
thermore,bloodglucosemaybemeasuredafterthechildhasgonetobed.Thiscanaf‐
fectthequalityofsleep,inparticularifthechildneedstoeatduetolowbloodglucose
level.
Timeofyouthisachallengingperiodoflife;adolescentswithdiabetesfaceuniqueage‐
specificdemands.Somedifficultiesarelearningaboutthenewdiseaseandmanaging
23 Background
diseaseknowledge,maintainingapositivehealthbehaviourandensuringtreatmentre‐
gimeadherence.Youngpeoplefacedifficultiesinthetransitionfromchildhoodto
adulthood,andproblemscopingwithachronicdiseasearecommon.Managingemo‐
tions,navigatingsocialrelationshipswithpeersincludingthedisclosureofthedisease
aresomeoftheirdifficulties(13).Aperiodofdiseaseneglectisoftenseen,resultingin
areducednumberofdailymeasurementsandpoordiabetescontrol.Thiswillleadto
severalyearswithelevatedbloodglucose,thusincreasingtheriskofdevelopingdiabe‐
tesrelatedcomplicationslaterinlife.
Managementofthecondition
Withoutproperdiseasemanagementtheindividualislikelytobecomeprogressivelyill
anddebilitated.Managementofbloodsugarlevels‐hypoglycaemiaandhyperglycae‐
mia‐isextremelyimportantindiabetescare.Ifbloodglucoseisuncontrolleditcan
leadtocomplicationssuchasretinopathy,nephropathy,neuropathy,orheartdisease,
strokeandperipheralvasculardisease.Hypoglycaemiaoccurswhenthelevelofblood
glucosefallsbelow3.5millimolesperlitre(mmol/L).Indicationsincludehunger,nerv‐
ousness,shakiness,perspiration,dizziness,sleepinessandconfusion,andifunattended
itmayleadtounconsciousness.Fearofrecurrenthypoglycaemiacandecreasequality
oflifeintheshortterm,butcanalsohinderadherencetotreatmentandtheachieve‐
mentofgoodglycaemiccontrol.Hyperglycaemiaoccurswhenthelevelofbloodglucose
ishigherthan11millimolesperlitre.Hyperglycemiatogetherwithinsulindepletion
canleadtoketoacidosis.Thisisafearedconditionthatuntreatedhasahighmortality.
Admissiontohospitalisnecessary,oftenatanintensivecareunit.Long‐termconse‐
quencesofchronichyperglycaemiamaybenervedamage,kidneydamageorfailure,
damagetothebloodvesselsoftheretinaandothereyecomplications.Prematuremor‐
bidity,mortality,reducedlifeexpectancyandfinancialandothercostsofdiabetesmake
itanimportantpublichealthcondition(14).
Glucoseisgenerallymeasuredinthreeways:
a)first,bloodglucosecanbetestedbyadropofbloodwithaglucosemeter(ca‐pillarybloodglucosetesting),alsoknownasself‐monitoringofbloodglucoseorSMBG;b)second,continuousglucosemonitors(CGM)providefrequentautomatedtest‐ingofinterstitialtissueglucose,calibratedtoreflectbloodplasmaglucose;andc)third,longertermcontrolismeasuredbyglycatedhaemoglobin(HbA1c)whichreflectstheaveragebloodglucoselevelsover2to3months.
24 Background
Psychosocialandculturalaspectsofthediseasearealsoimportantconsiderationsin
successfuldiabetesmanagement.Positivefactors,e.g.,goodcopingskills,familysup‐
port,effectiveweightcontrolprograms,etc.,canincreaseinterestindiseasemanage‐
mentandimproveadherencetomedication,resultinginbetterglycaemiccontroland
improvedqualityoflife.Individualslackingpositiveinfluencesmayfinddiseaseman‐
agementmoredifficult,resultinginpoorglycaemiccontrolandanincreasedlikelihood
forlong‐termcomplications(15;16).
Self‐monitoringofbloodglucoseisanessentialpartofdiabetesmanagementandis
usedtooptimiseglycaemiccontrol.Managementofdiabetesrequireslifelongintegra‐
tionofmanyfactorssuchas,lifecircumstances,dailyadherencetodietaryandexercise
plans,frequentbloodglucosemonitoringandadherencetomedications.Trainingfor
self‐managementimprovestheknowledgeofdiabetes,glucoselevelsandglycatedhae‐
moglobin.Itcanalsoleadtoimprovedsystolicbloodpressurelevels,bodyweightand
canreducetheneedforDMmedicationfortype2DM.Effectivecontrolofbloodglu‐
coselevelsallowsindividualswithDMtoadjusttherapy(insulindosage)appropriately.
Impairedawarenessofhypoglycaemia
Recognizingsymptomsofhypoglycaemia(being‘aware’)attheironsetisfundamental
fortimelyself‐managementofbloodsugarlevels.Impairedawarenessofhypoglycae‐
miameanstheindividual’sabilitytoperceivetheonsetofhypoglycaemia,ortorecog‐
nizewarningsymptoms,isdiminishedorabsent.Thecounterregulatoryhormonere‐
sponse(insulin)isdeficientorlackinginindividualswithimpairedawareness(17).Im‐
pairmentintheabilitytorecognizetheonsetofhypoglycaemiacanhaveseriousconse‐
quencesandconstitutesasignificantproblemcommonlyseeninindividualswithtype
1DM,andlessoftenseeninindividualswithtype2DM.CGMisausefulsystemthat
helpsindividualstodetectasymptomatichypoglycaemia,asitprovidesreal‐datatime
andalertsfortheindividual(18).Hypoglycaemiaisthemaincauseofindividualsget‐
tingacontinuousglucosemonitoringsysteminNorwaytoday.
Norwayimplementedanationaldiabetesstrategy(2006‐2010laterprolongedto
2012)whichaimedtoimprovetheprimarypreventionofdiabetes,improvethecoop‐
erationbetweentheprimaryandsecondaryhealthcareandincreasetheresourcesof
healthcareinthelocalmunicipalities.Norwaypublishedthelatestnationaldiabetes
clinicalguidelinein2016.Theclinicalguidelinehasitsmainfocusonsecondaryand
tertiarypreventionofdiabetes.Althoughbothdocumentsemphasiselifestyleandself‐
managementofthedisease,theuseofFreeStyleLibreisnotincluded(19).
25 Clinical effectiveness and safety
Clinicaleffectivenessandsafety
Issueaddressed
Accordingtothesubmitter,byusingFreeStyleLibre,peoplewithDMmayimproveglu‐
cosecontrolmanagementandconsequentlyreducethenumberofdiabetesrelated
complicationsandimprovetheirqualityoflife.Moreover,thetechnologycouldmakeit
easierforpeopletoadheretotreatment.FreeStyleLibremayhelpreducingtheinci‐
denceofsevereandnocturnalhypoglycaemiaanditsassociatedanxiety.FreeStyleLi‐
bremayofferbenefitsthroughcostandresourcesavingsbyreducingthenumberof
hospitaladmissionsandconsultationsfordiabetesrelatedcomplications,andby
achievingoptimalbloodglucoselevelsmorequickly.
Objective
Theobjectiveofthisdocumentwastoassesstheclinicaleffectiveness,cost‐effective‐
nessandsafetyofFreeStyleLibreforinsulintreatedindividualswithtype1and2DM
(“type1and2”).
Methods
Inclusion and exclusion criteria
Weusedthepopulation,intervention,comparison,outcome,anddesign(PICO–D)
frameworktoevaluatethesuitabilityofevidenceforinclusion(seeTable1).
WeselectedtheseoutcomesincollaborationwiththeclinicalexpertsandtheNorwe‐
gianDiabetesAssociationviacommunicationandconsultationmechanisms.
26 Clinical effectiveness and safety
Table1.PICO–Dframework
Population Insulin dependent individuals (of any age) diagnosed with diabetes type 1 or 2
Intervention FreeStyle Libre flash glucose monitor
Comparator Any other glucose monitoring system or procedure including conventional self-monitoring of blood glucose plus multiple insulin injections, pen device use, insulin syringe, etc.
Outcomes Change in HbA1c (glycosylated haemoglobin),
Hypoglycemia or hyperglycemia – day, night time, and episodes
Quality of life
Patient / treatment satisfaction
Pain
Adverse events (related to the device or not, withdrawals, etc.)
Design Randomised control trials (RCTs)* or controlled studies (i.e., controlled before and after with at least two intervention and two control sites; or interrupted time series with at least three data points before and three after the time point of the intervention).
* Hierarchies of evidence rank research according to its validity. RCTs are commonly viewed as providing the highest level of evidence. This type of design minimises the risk of confounding factors influencing the results and it has been the ‘gold stand-ard’, or optimal research design, for evaluating effectiveness. Weexcludedstudiesif:
Populationofinterest/focusofthepublicationwasotherthaninsulindependent
diabetestype1and2DM
Theglucosemonitoringsystemintheinterventionwasdonebyaninstrument
otherthanFreeStyleLibre
NoneoftheoutcomesinTable1wereassessed
Datafortype1and2DMwerenotpresented/availableindependently
Typeofstudywereclinicalpracticeguidelines,conferenceabstractsandproceed‐
ings,books,bookchapters,animalormodellingstudies
Literature search and selection of studies
Thesearchstrategywasdesigned,peerreviewedandexecutedbytwoexperiencedin‐
formationspecialists.Thesearchwasadaptedtoeachdatabaseandhadnolanguage
restrictions.Wesearchedsystematicallyinthefollowingdatabases:MEDLINEandEm‐
baseviatheOvidinterface,CochraneLibrary:CochraneDatabaseofSystematicRe‐
views,OtherReviews,CentralRegisterofControlledTrials,EconomicEvaluationsdata‐
base,CentreforReviewsandDissemination:DatabaseofAbstractsofReviewsofEf‐
fects,HealthTechnologyAssessmentdatabaseuptoJanuary18,2017.Wealso
searchedtheTheInternationalNetworkofAgenciesforHealthTechnologyAssessment
(INAHTA)members’websiteforotherFreeStyleLibreassessment.
27 Clinical effectiveness and safety
WesupplementedthedatabasesearchbyconsultingothersourcessuchasGoogle
scholar,HTAagencies’homepages,ClinicalTrials.gov,andWHOInternationalClinical
TrialsRegistryPlatformandbyhand‐searchingreferencelistsofrelevantpapers.We
usedacombinationofsubjecttermsandtextwords(seeAppendix2fordetailedsearch
strategy).
Pairofauthors(JB,KF,TTorIS)independentlyscannedtitleandabstractofthere‐
trievedrecords.Weinvestigatedallpotentiallyrelevantarticlesasfulltextregarding
eligibilitycomplianceaspercriteriamentionedinTable1.Differencesinopinionwere
resolvedbyathirdreviewer.TheresultsofthisprocessarepresentedinaPRISMA
(preferredreportingitemsforsystematicreviewsandmeta‐analysis)flowchart(Fig‐
ure3intheresultssection).
Alldatawasextractedindependentlybyonereviewer(JB)intoastandardiseddataex‐
tractionform,whichwasthenbeencheckedforaccuracybyanotherreviewer(KFor
TT).Weextractedthefollowingdata(seeTable2):
Table2.Dataextracted
Data Details to be extracted (if available)
Publication Summary Author & year, title, publication type, inclusion/exclusion, country (use first author affiliation) of origin
Population Total sample size, age, gender, diagnosis, years since diagnosis, type of di-abetes
Intervention Length, follow up
Comparator Blood capillary glucose monitoring, other system (Navigator, MiniMed) etc.
Outcomes Same as in table 1
Weextracteddatafromincludedstudiesasfaraspossible.Wecontactedprimarystud‐
ies’authorstorequestdataprovidedingraphicalformfortwooftheoutcomes.Pri‐
maryauthorssentourrequesttothesubmitter;duetoadelayintheresponse,weesti‐
mateddatafromthegraphusingagraphicaldatasoftware(EngaugeDigitizer).We
eventuallyreceivedaresponsefromthesubmitter.
Risk of bias assessment in included studies
Tworeviewers(JBandKF)independentlyevaluatedtheriskofbiasusingtheCochrane
"Riskofbiastool”(20).Theriskofbiastooladdressesspecificdomainsrelatedtothe
study’sinternalvalidity:sequencegenerationandallocationconcealment(selection
28 Clinical effectiveness and safety
bias),blindingofparticipants,studypersonnelorstaff(performancebias),blindingof
studyassessors(detectionbias),incompleteoutcomedata(attritionbias),selective
outcomereporting(reportingbias),andothersourcesofbias.Abriefdescriptionof
thesebiasesispresentedinTable3below.
Table3:AssessmentofRiskofBiasinincludedRCTs
Domain-Item Description
Sequence Generation Was the allocation sequence adequately generated?
The method used to generate the allocation se-
quence should be described in sufficient detail
to allow an assessment of whether it should
produce comparable groups.
Allocation Concealment Was allocation adequately concealed?
The method used to conceal the allocation se-
quence should be described in sufficient detail
to determine whether intervention allocations
could have been foreseen in advance or during
enrolment.
Blinding of participants, personnel and outcome asses-sors Was knowledge of the allocated intervention adequately prevented during the study? Assessments will be made for each main outcome (or class
of outcomes).
All measures used, if any, to blind study partici-
pants and personnel from knowledge of which
intervention a participant received, should be
described. Any information relating to whether
the intended blinding was effective should also
be reported.
Incomplete outcome data Was incomplete outcome data adequately addressed? Assessments will be made for each main outcome (or class
of outcomes).
The completeness of outcome data for each
main outcome should be described, including
attrition and exclusions from the analysis. The
authors should report any attrition and exclu-
sions, the numbers in each intervention group
(compared with total randomized participants),
and reasons for attrition/exclusions and any re-
inclusions in analyses.
Selective Reporting* Were results of some outcomes not reported because the results were not statistically significant Assessment will be made by comparing trial registry record
or published protocol to the study publication
The authors should not select and report out-
comes in full text publications based on their
results. All of the study’s prespecified out-
comes should be reported (by checking the trial
protocol if available)
29 Clinical effectiveness and safety
*Betweenstudiesreportingbiases:duetoalackofsamplestudies(i.e.,morethan10)
wewerenotabletoproducedfunnelplots(20)toinvestigatepublicationreporting
bias.
Withinstudiesreportingbiases:whenapublishedstudyprotocolwasavailable,we
comparedoutcomesinthestudyprotocolwiththeoutcomesinthepublishedreport.
Wedocumentedthetrialnumberortheavailabilityofapublishedprotocol.
Eachcriterionwasratedasbeingatlow,high,orunclearriskofbiasaccordingtothe
informationprovidedinthestudies.Weclassifiedstudiesas'lowriskofbias'ifallkey
domainshadlowriskofbiasandnoseriousflaws.Thecriterion'unclearrisk'wasas‐
signedwhentheabsenceorambiguityoftheinformationhinderedtheassessors'abil‐
itytodeterminethepotentialforbias.Ifthecriterionwasnotfulfilledweclassifieditas
‘highrisk’.Disagreementswereresolvedthroughconsensusmeetingsbyconsultinga
thirdteammember.
Theresultsoftheriskofbiasassessmentswereusedfordescriptivepurposestopro‐
videanevaluationoftheoverallqualityoftheincludedstudiesandatransparent
methodofdevisingrecommendationsforthedesignoffuturestudies.
Data synthesis
Whentwoormorestudiesreportedthesameoutcome,wepooledthedata(meta‐anal‐
ysis)usingRevMan5.32014(21).Forcontinuousdata,weusedthegrouppost‐test
meansandstandarddeviationstocalculatetheeffectsize.Effectsizeswereexpressed
preferentiallyintheformofmeandifference(MD)and95%confidenceinterval(CI).
Dichotomousdatawasanalysedbycalculatingrelativerisk(RR)andthecorresponding
95%CI.Forgraphicaldata,standarderrorsweretransformedtostandarddeviations.
Weorganisedthedatabyoutcomeandreportedtheresultsfortheoutcomesofinter‐
estintextandtables.
Assessment of Heterogeneity
Statisticalheterogeneitywasassessedbyvisualinspectionoftheforestplottoassess
obviousdifferencesinresultsbetweenthestudies,andusingtheI²andchi²statistical
Other sources of bias Was the study apparently free of other problems that could put it at a high risk of bias?
Overall, the study should be free from any im-
portant concerns about bias (i.e. bias from
other sources not previously addressed by the
other items).
30 Clinical effectiveness and safety
tests.AsrecommendedintheCochraneHandbookforSystematicReviewsofInterven‐
tions(22),wefollowedtheinterpretationofanI²valuefrom0%to40%as'mightnot
beimportant';from30%to60%asrepresenting'moderate’heterogeneity;from50%
to90%asrepresenting'substantial'heterogeneity;andfrom75%to100%asrepre‐
senting'considerable'heterogeneity.BecauseI²hasoverlappingcategories(i.e.,0%to
40%,30%to60%)or"ambiguous"zones,whenmoderatetosubstantialstatisticalhet‐
erogeneitywasfound(i.e.,I²between50%and60%)weexploreditthoroughly.Inad‐
dition,clinicalandmethodologicaldiversitywasassessedintermsofparticipants,out‐
comes,andstudycharacteristicstodeterminewhetherameta‐analysiswasappropri‐
ate.
Giventhatvaluesbetween50%and60%fallinan'ambiguous'zone,whentherewere
noapparentcausesofheterogeneity,wekeptthetrialintheanalysisanddocumented
ourdecision.Thechi²testwasinterpretedwithapvalue≤0.10indicatingevidenceof
statisticalheterogeneity.
Grading of the evidence
Onereviewer(JB)usedtheGRADEtool(GradingofRecommendationsAssessment,
DevelopmentandEvaluation)developedbytheGRADEworkinggroup(23)todeter‐
minethecertaintyoftheeffectsofinterventionsreportedintheincludedreviews,i.e.
towhatdegreewecouldtrusttheresults.Asecondreviewer(KF)independently
checkedtheassessment.Ifdisagreementswerefound,theyweresolvedbydiscussion.
Weconsideredthecompileddocumentationforeachofthemainoutcomesusing
GRADEandpreparedsummaryoffindingstablesfortheoutcomesofinterest.Intheta‐
blesweintegratedthecertaintyofevidenceandthemagnitudeofeffectoftheinterven‐
tion.WemadetheGRADEratingsseparatelyforeachoftheoutcomesofinterest.
WeusedthefiveGRADEconsiderationsfordowngrading(studylimitations,con‐
sistencyofeffect,imprecision,indirectnessandpublicationbias)toassessthecertainty
ofthebodyofevidenceashigh,moderate,loworverylow.Thetablebelow(seeTable
4)presentswhatGRADEmeansbyeachofthesefourcategories.
Evidencefromrandomisedcontrolledtrials(RCTs)startedashighqualityevidencebut
mayhavebeendowngradeddependingonthefivecriteriainGRADEthatareusedto
determinethecertaintyoftheevidence.
31 Clinical effectiveness and safety
Table4.GRADE
QualityLevel Significance
High We are very confident that the true effect lies close to that of the estimate of the effect
Moderate We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low Our confidence in the effect estimate is limited: The true effect may be substan-tially different from the estimate of the effect
Very low We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
Presentation of results in relation to the submitter’s package
Someresults(i.e.,searchandriskofbias)fromthesubmitter’sandtheNIPHhavebeen
comparedintabularform.Inthediscussionchapter,weincorporatetheclinicaleffec‐
tivenessimplicationsoftheresultsandpresentagreementsordisagreementswiththe
submitter’sfile.
Stakeholder involvement
Asstatedintheliterature,thereisnotaconventionalmodeltoinvolvestakeholders’in
HTAs(24).Nevertheless,theNIPHfollowedaconsultationprocessaimingtoincorpo‐
ratetheexperienceandknowledgeofclinicalexperts,consumersorusersofthedevice,
andpatientorganizations.Clinicalandexternalreviewexpertsweredesignatedbythe
RegionalHealthAuthorities.
First,thePICOframeworkwasagreeduponincollaborationandwiththefeedback
fromclinicalexpertsandtheNorwegianDiabetesAssociaton.Second,weaskedclinical
expertstoprovidefeedbackduringtheassessmenttovalidateourunderstandingofthe
diseaseandinitialfindings.Third,allstakeholders(clinicalexperts,users’andpatients’
organization)wereinvitedtoprovidefeedbackonthefirstdraftoftheassessment.The
projectleadercontactedthemandprovidedinitialinformation.Followinganinvitation
totakepartintheassessment,everyoneagreedtotheconfidentialitytermsandcondi‐
tionsandsignedcorrepondingforms.Users’expertswereinvitedfollowingasnowball‐
ingtechnique,andtheNorwegianDiabetesAssociation,themainstakeholderinthisre‐
port,wassentanemailinvitation.TheNIPHrequestedclarificationwhenneededon
thekeyevidenceresponsessubmitted.Lastly,followingthefirstdraft,weincorporated
feedbackintothemanuscriptandprepareditforexternalcontentexperts.Then,meth‐
odologistsatNIPHandexternalcontentexpertspeer‐reviewedthesecondandfinal
draft.
32 Clinical effectiveness and safety
Results
Literaturesearch
Thesearchidentified1665citations,whichincluded1622journalrecordsand43trial
registryrecords.Weexcluded1576recordsbasedonscreeningoftitlesandabstracts.
Weassessed89fulltextarticlesandtrialregistryrecordsforeligibility.Thisresultedin
thefinalinclusionof2RCTsandcorrespondingregistryrecords.SeeFigure3forde‐
tails.
Severalsingle‐armstudies,whichprovideevidenceofinitialexplorationsofFreeStyle
Libreaccuracy,comparedsensorreadingstobloodglucosecapillarymeasurementsor
venousandcapillary‐pairedmeasurements.Giventheavailabilityofrandomizedcon‐
troltrialsandlackofacomparisongrouponthistypeofdesign,theteamdecidedtofo‐
cusontheevidenceprovidedbytheRCTs.Asummaryofcompleted(publishedandun‐
published)single‐armstudies(e.g.,populationstudies,author&year,countryoforigin,
duration,andoutcomes)ispresentedinAppendix3.
FreeStyleLibreregistryrecords
WefoundseveralregistryrecordsforFreeStyleLibre.Amongthese,sevensinglearm
studieshavebeencompleted,threeareongoingRCTs,andsevenareongoingsingle
armstudies.Wereporteddetails(e.g.,registrynumber,studytitle,studystatus,coun‐
tryoforigin,outcomes,andstudyduration)ofongoingstudiesinAppendix4.
OtherEuropeanassessments
WefoundfivecompletedandoneongoingFreeStyleLibreassessments.Also,wefound
aGovernmentclaim,apeadiatricguidelineandaSwedishreportusingFreeStyleLibre
asanexample.Wepresenttheresultsofthesereportsinthediscussion.
33 Clinical effectiveness and safety
Figure3.Aflowchartoftheliteratureselection
Table5(below)presentsthecomparisonbetweenthesearchstrategyofNIPHandthe
submitter:thenumberofdatabasessearcheddifferedgreately,andsowerethe
numberoffulltextretrievedandscreened,otherHTAEuropeanassessments,and
ongoingstudiesandrecordsincluded.
Table5.NIPHandsubmittersearchfindings
Norwegian Institute of Public Health Submitter
Database search: MEDLINE, Embase (Ovid), Cochrane Library, Technology Assessments, Economic Evaluations database, Centre for Reviews and Disseminations, Database of Abstracts of Reviews of Effects, Health Technology Assessment, and PubMed databases up to January 18, 2017.
Database search: (pg 11) www.ncbi.nlm.nih.gov/gov/pubmed www.cochranelibrary.com up to Nov 11, 2016
Full text records screened: 89
Full text records retrieved: 4
Records identified through systematic searching
(n = 1622)
Additional records identified through other sources
(n = 43)
Total records screened (n = 1665)
Excluded records (n = 1576)
Full text assessed for eligi‐bility
(n = 89)
Full text records excluded with reasons (n = 85)
Full text publications included n=2 (and 2 registry records)
34 Clinical effectiveness and safety
Reports from other HTA agencies: 1 ongoing and 4 published
Reports from other HTA agencies: 1
Registry records: 18
Ongoing: “to be supplied”
Records included in the analysis: 2 RCTs and 2 registry records
Records included: 1 article and 1 abstract/poster
Characteristicsofincludedtrials
Included Studies
TwoRCTs(andcorrespondingregistryrecords–hereafter“protocols”)wereincluded
inthisappraisal.Thestudieswerepublishedin2016and2017,andwereconducted
across23(25)and26Europeandiabetescentres(26)(IMPACTandREPLACEstudies
respectively)insevencountries(France,Germany,UnitedKingdom,Sweden,Austria,
SpainandtheNetherlands).Inbothtrials,thelengthofthetreatmentwas6months,
thetrialsusedarandomisedclinicaltrialwithparallelgroupstudydesign,andthe
FreeStyleLibregroupwascomparedtoSMBG.Individuals’insulinadministration
methodwasprimarilymultipledailyinjections(25;26).
Participants
Thestudiesincludedatotalof463adults,ofreportedwhiteethnicity(177females,286
males).DetailsofinclusionandexclusioncriteriatothetrialscanbeseeninAppendix
5.Table6showssomeimportantbaselinecharacteristics.TheIMPACTstudyincluded
individualswithtype1DMwhoshowedgoodglycaemiccontrolatbaseline,whilethe
REPLACEstudyincludedindividualswithtype2DMwithbaddiabetescontrol,HbA1c
of8.74%.
Table6Baselinecharacteristicsofincludedstudies
Study (Diabetes type)
BMI (SD) Age (range) Baseline HbA1c % (SD)
Years since diagnosis
SMBG per day (SD)
Bolinder (25):
Type 1 DM
25 (4) 43 years
(33-57)
6.7 (0.6) 20 years
(range 12-31),
5.5 (2.2)
Haak (26):
Type 2 DM
33 (6) 59 years
(22-81)
8.74 (1.04) 17 years
(range 2-43)
3.8 (1.3)
BMI: body mass index; DM: diabetes mellitus; SMBG: self monitor blood glucose; SD: standard deviation;
FreeStyle Libre vs SMBG
SMBGincludesselfmonitoringandthefollowupuseofinsulinintheformofapende‐
vice,orcontinuoussubcutaneousinsulininfusion.Inbothstudiesparticipantswore
35 Clinical effectiveness and safety
FreeStyleLibretechnologyintomaskedmodefor14daysbaselineperiod;measure‐
mentswereblindedforparticipantandinvestigatoratthistime.Participantssup‐
portedtheirglucosemanagementbySMBG.TheyusedthestripportbuiltintotheFree‐
StyleLibreandcompatibleteststripsmadebytheAbbott.Participantswereaskedto
keeprecordofcapillaryglucoseconcentrationsinaglucosediaryandtologother
eventsinaneventdiary.Thosethathadsensordatafor50%oftheblindedwearpe‐
riod,ormorethan650individualsensorreadings,werethencentrallyrandomisedinto
twogroups.
Afterrandomisation,thetechnologywasunblindedforparticipantsintheintervention
groupwhothencontinuouslyusedsensorglucosedataforself‐managementofglucose
levelsthroughoutthedurationofthestudy.Participantsintheinterventiongroupwere
givenaccesstothetechnologysoftware,whichtheycoulduseathometoreviewtheir
sensordataiftheywished.Notrainingwasprovidedtotheseparticipantsforinterpre‐
tationofglucosesensordatainneitherofthestudies.Participantsinthecontrolgroup
self‐monitoredglucoseconcentrationsusingtheFreeStyleLibremeterandteststrips.
Figure4.Studydesign(reprintedfromBolinder2016withsubmitter’spermission)
Riskofbiasinincludedstudies
Tworeviewersevaluatedthe‘riskofbias’usingtheCochraneHandbooktool(20)
basedontheprimaryarticleandcorrespondingprotocol.Resultsoftheriskofbiasfor
thetwoRCTsarepresentedbelowinFigures5and6.Someconcernsintheriskofbias
weretheunclearinformationaboutallocationconcealment–althoughtherewereno
obviousbaselineimbalancesnoinformationisprovidedonconcealmentallocation‐
andlackofblindingintheincludedstudies(participants,personnel,andstaff).Weun‐
derstandthatforparticipantsandpersonnelitisalmostimpossibletoperformatrial
withtrueblindingwiththistypeofintervention.However,thisdoesnotmeanthatpo‐
tentialbiasescanbeignored.Thefactthatparticipantsandpersonnelwerenotblinded
36 Clinical effectiveness and safety
maybiastheresultsandoutcomeassessment.
Detectionbiasforsubjectiveoutcomes(i.e.,qualityoflife,treatmentsatisfaction)was
assessedashighrisk,asitispossiblethatthenon‐blindedassessor(theindividual)
overestimatedtheeffectoftheintervention.Detectionbiasforsomeobjectiveout‐
comes(i.e.adverseevent)weretheobserverjudgementwasinvolvedwasassessedas
highrisk;foroutcomeswerenojudgementwasinvolved(i.e.,bloodtest)weconsid‐
eredtheassessmentoftheoutcomewasnotlikelytobeinfluencedbythelackofblind‐
ing.Furthermore,weareunclearhowthedatawastransferredandanalysedfromthe
FreeStyleLibreforanalysis.Itisclear,however,thatallstudypersonnelwasunmasked
andthereforethenon‐blindedassessorcouldoverestimatetheeffectbeingassessed.
Selectiveoutcomereportingwasassessedasunclear,astherearedifferencesbetween
theprotocol,thejournalpublication,andthesubmitterpackage,whichwecouldnot
explain.Overall,ourassessmenthadahighernumberofhighriskandunclearrisksdo‐
mainsthanthesubmitter’sassessment.Table7belowcomparesresultsregardingrisk
ofbias:
Table7.NorwegianInstituteofPublicHealthandsubmitter’sriskofbias
Norwegian Institute of Public Health Submitter
Tool utilized: Cochrane Handbook Tool (20)
Selection bias (randomization): Low risk
Selection bias (allocation): Unclear risk
No information on how this was done
Performance bias (blinding of personnel and
participants): High risk – no blinding
Detection bias (subjective outcomes): High risk
– self reported
Detection bias (objective outcomes): low to high
risk – outcomes requiring judgement high risk,
outcomes not requiring judgement are low risk
Attrition bias: Low risk – data accounted for, ap-
propriate statistical methods
Tool utilized: no information provided
Selection bias (randomization & allocation): Low
risk
Central interactive web response WRS using the
biased coin minimisation method; study centre
and type of insulin administration were prognostic
factors
Performance bias: Unclear risk -
no blinding
Detection bias: Low risk
Detection of hypoglecemia was done by sensor
and not subject to human evaluation
Attrition bias: Low risk
Minimal missing data. Missing data is accounted
for
37 Clinical effectiveness and safety
Reporting bias: Unclear risk
Other sources of bias: High risk -sponsor stud-
ies
Reporting bias: Low risk
All outcomes accounted for
Other sources of bias: not evaluated
Figure5.RiskofBias:BolinderandHaak
Figure6.RiskofBias:studiescombined
38 Clinical effectiveness and safety
EffectofFreeStyleLibre
Theincludedstudiesprovideddataforhealthrelatedqualityoflife,treatmentsatisfac‐
tion,HbA1c,glycaemicmeasures,andadverseevents.Thestudiesdidnotassesspain
asaseparateoutcome,butrecordedpainaspartofadverseevents.TheRCTsspecified
theevaluationoftheself‐managementofDMwasinthehomesetting.Theresultsre‐
latedtotheeffectsoftheinterventionfromthemeta‐analysisarepresentedbelowcor‐
respondingtotheobjectivesoftheassessment.
Followingthefeedbackfromclinicalexperts,wepresenttheeffectestimatesofeach
groupi.e.type1DMandtype2DMfortheglycaemicmeasuresforeachofthetwo
studiesseparatelyintheforestplotsandinAppendix6.Theforestplotsarefollowed
bythe“summaryoffindings”table,whichisthekeyinformationconcerningthecer‐
tainty(quality)ofevidenceandthemagnitudeofeffectoftheintervention.
Outcomes
Alloutcomes,withtheexceptionofadverseevents,weremeasuredatbaseline
and6monthsendpoints.Adverseeventsweremeasuredatbaselineand6
monthsendpointfortheinterventiongroup,andbaselineand4weeks,divided
intwosetsof14daysat3monthsand6month,inthecontrolgroup.
Health Related Quality of Life
Thisoutcomewasmeasuredwithadiabetesspecificscale(self‐report,DiabetesDis‐
easeQuestionnaire);thescaleisa1‐5pointLikertscalewithhighscoresindicatingdis‐
satisfaction,frequentimpact,orfrequentworry.TheMDwas‐0.05(95%CI‐0.16to
0.05;I20%indicatingnoheterogeneity;p=0.36).
39 Clinical effectiveness and safety
TheGRADEqualityofevidenceforthisoutcomewaslowwhichmeansFreeStyleLibre
mayleadtolittleornodifferenceinhealthrelatedqualityoflifeforindividualswith
type1or2DM.
Table8.Summaryoffindingstable‐healthrelatedqualityoflifeFreeStyle Libre compared to self-monitoring of blood glucose in insulin-treated diabetes individuals
Patient or population: individuals with diabetes type 1 and 2 insulin dependent Setting: home setting Intervention: FreeStyle Libre technology Comparison: SMBG Outcome: measured at 6 months (end of intervention)
Outcomes Anticipated absolute effects* (95% CI) Relative effect
(95% CI)
№ of parti-
cipants
(studies)
Quality of the
evidence
(GRADE)
Risk with SMBG Risk with FreeStyle Libre technology
Health Related Quality of Life DQoL Likert scale (self-reported) from 1 to 5, lower scores are best
The mean health related quality of life ranged across control group from 2 to 2.2 points
The mean health re-lated quality of life in the intervention group was 0.05 lower (0.16 lower to 0.05 higher)
- 465 (2 RCTs)
⨁⨁◯◯ LOW a
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RCT: randomised control trial:
a. Issues related to unclear or high risk of selection, performance and detection biases (subjective outcome) (downgraded twice) Patient Treatment Satisfaction
Thisoutcomewasmeasuredwiththe(self‐report)DiabetesTreatmentSatisfaction
Questionnaire;scoresrangefrom‐18to18withhigherscoresindicatinghighertreat‐
mentsatisfaction.TheMDwas5.10(95%CI2.95to7.26;I2=70%indicatingsubstan‐
tialheterogeneity;P=0.07alsoindicatingstatisticallysignificantheterogeneity).
Duetostatisticalheterogeneity(I2=70%),clinicalheterogeneitywasexplored.On
evaluation,dailylifeofindividualswithType1and2DMisverydifferent.Buteven
whendiabetesmanagementimposesconsiderabledemandsonindividuals,treatment
methodusedhasanimpactontreatmentsatisfaction.Despitehighheterogeneityboth
groupsconsiderablyimprovedtheirtreatmentsatisfactionattheendofintervention.
40 Clinical effectiveness and safety
TheGRADEqualityofevidence(seetable9)forthisoutcomewaslow,whichindicates
thatFreeStyleLibremayimprovetreatmentsatisfactionforindividualswithtype1or
2DM.
Table9.Summaryoffindings‐patienttreatmentsatisfactionFreeStyle Libre compared to self-monitoring of blood glucose in insulin-treated diabetes individuals
Patient or population: individuals with diabetes type 1 and 2 insulin dependent Setting: home setting Intervention: FreeStyle Libre technology Comparison: SMBG Outcome: measured at 6 months (end of intervention)
Outcomes Anticipated absolute effects* (95% CI) Rela-tive effect (95% CI)
№ of partici-pants
(studies)
Quality of the evidence (GRADE)
Risk with SMBG
Risk with FreeStyle Libre technology
Patient satisfaction Satisfaction score (self-reported), scale from: -18 to 18, high scores are best
The mean pa-tient satisfac-tion ranged across control groups from 7 to 9 points
The mean patient sat-isfaction in the inter-vention group was 5.1 higher (2.95 higher to 7.26 higher)
- 465 (2 RCTs)
⨁⨁◯◯ LOW a
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RCT: randomised control trial:
Sponsored studies but not further downgraded a. Issues related to unclear or high risk of selection, performance and detection biases (subjective outcome) (downgraded twice)
Pain: Painwasrecordedamongadverseevents
HbA1c:
HbA1c:(targetlevelof7%orless,analysedbyICONLaboratories,Dublin,Ireland).The
hemoglobinA1cMDwas‐0.00(95%CI‐0.14to0.14;I2=0%indicatingnoheterogene‐
ity;P=0.81alsoindicatingnostatisticallysignificantheterogeneity).
41 Clinical effectiveness and safety
TheGRADEqualityofevidence(seetable10)forthisoutcomewaslow,whichindicates
thatFreeStyleLibremayleadtolittleornodifferenceonHbA1cchangesforindividuals
withtype1or2DM.
Table10.Summaryoffindingstable‐HbA1c%FreeStyle Libre compared to self-monitoring of blood glucose in insulin-treated diabetes individuals
Patient or population: individuals with diabetes type 1 and 2 insulin dependent Setting: home setting Intervention: FreeStyle Libre technology Comparison: SMBG Outcome: measured at 6 months (end of intervention)
Outcomes Anticipated absolute effects* (95% CI) Relative effect (95% CI)
№ of partici-pants
(studies)
Quality of the evidence (GRADE) Risk with SMBG Risk with FreeStyle
Libre technology
HbA1c Target level 7% or less
The mean HbA1c (%) ranged across control groups from 7 to 8%
The mean HbA1c (%), target level 7% or less in the inter-vention group was 0 (0.14 lower to 0.14 higher)
- 462 (2 RCTs)
⨁⨁◯◯ LOW b
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RCT: randomised control trial:
b. Issues related to unclear or high risk of selection and performance biases (downgraded twice)
Glycaemic Measures (time in range, without hypoglycaemia and hyper-glycaemia)
Timeinrange‐glucose3.9‐10mmol/L
Thetime(inhours)spentwithglucoserange3.9‐10mmol/L,showsastatisticallysig‐
nificantresultinfavorofFreeStyleLibre.TheoverallMDwas1.00(95%CI0.32to
1.68;I2=6%indicatingnoheterogeneity;p=0.30).Resultsshowedstatisticallysignifi‐
canteffectfortype1DM(238participants,MD1.2095%CI0.46to1.94)andnotsta‐
tisticallysignificanteffectfordiabetestype2(224participants,MD0.40,95%CI‐0.93
to1.73).TheGRADEqualityofevidenceforthisoutcomewaslow(seeTable11)which
meansFreeStyleLibremayslightlyimprovetimewithglucoseinrange3.9‐10ml/Lfor
insulindependentindividualswithtype1and2DM.
42 Clinical effectiveness and safety
Glucose<3.9mmol/Lwithin24h
Time/hours
Thetime(inhours)spentwithglucose<3.9mmol/Lwithin24hoursdecreasedinthe
FreeStyleLibregroup.ThetotalMDwas‐0.60(95%CI‐0.88to‐0.32;I2=84%indicat‐
ingconsiderableheterogeneity,p=0.01).Thehighheterogeneityinglycaemicmeasures
isnotsurprisingandcanbeexplainedbyconsideringthephysiopathologyofeachtype
ofdiabetes.TheGRADEqualityofevidenceforthisoutcomewasverylow(seeTable
11)whichmeansweareuncertainwhetherFreeStyleLibreleadstolesstimewithglu‐
cose<3.9mmol/Lwithin24hforindividualswithtype1or2DM
Events
Themeannumberofeventswithglucose<3.9mmol/Ldecreasedforparticipantsinthe
intervention.ThetotalMDwas‐0.23(95%CI‐035to‐0.10;I2=64%indicatingsub‐
stantialheterogeneity;p=0.09).TheGRADEqualityofevidenceforthisoutcomewas
verylow(seeTable11)whichmeansweareuncertainwhetherFreeStyleLibrede‐
creasesthenumberofeventswithglucose<3.9mmol/Lwithin24hforindividualswith
type1or2DM
43 Clinical effectiveness and safety
Glucose<3.1mmol/Latnight(23:00‐06:00)within7hTime/hours
Thetimespentinhypoglycemiaduringthenightshowedanon‐statisticallysignificant
result.TheoverallMDwas‐0.22(95%CI‐0.46to0.03;I2=85%indicatingconsiderable
heterogeneity;p=0.010).Results,however,showedstatisticallysignificanteffectfor
type1DM(MD‐0.3595%CI‐0.51to‐0.19)andstatisticallysignificanteffect,however
smaller,fortype2DM(MD‐0.1095%CI‐0.20to0.00).TheGRADEqualityofevidence
forthisoutcomewasverylow(seeTable11)whichmeansweareuncertainwhether
FreeStyleLibredecreasesnocturnaltimewithglucose<3.1mmol/Lwithin7h
Events
Thenumberofhypoglycemicnighteventsshowedastatisticallysignificantdecreaseininfavouroftheintervention.TheMDwas‐0.09(95%CI‐0.13to‐0.04,I2=0%indicat‐ingnoheterogeneity;p=0.35).TheGRADEqualityofevidenceforthisoutcomewaslow(seeTable11)whichmeansFreeStyleLibremayslightlydecreasenocturnaleventswithglucose<3.1mmol/Lwithin7h
44 Clinical effectiveness and safety
Glucose <3.1 mmol/L within 24h
Time/hours
Thetimespentwithglucose<3.1mmol/Lwithin24hwasnotstatisticallysignificant.
ThepooledMDwas‐0.49(95%CI‐0.35to0.16;I2=89%indicatingconsiderableheter‐
ogeneity,p=0.002).Resultshowever,showedstatisticallysignificanteffectfortype1
DM(MD‐0.8595%CI‐1.24to‐0‐46)andhoweversmallerresults,theyarealsostatis‐
ticallysignificantfordiabetestype2(MD‐01895%CI‐0.35to‐0.01).TheGRADEqual‐
ityofevidenceforthisoutcomewasverylow(seeTable11)whichmeansweareun‐
certainwhetherFreeStyleLibreleadstolesstimewithglucose<3.1mmol/Lwithin
7hours.
Events
Similarly,resultsofthemeta‐analysisshowednon‐statisticallysignificanteffectsizein
thenumberofeventsparticipantshadwithglucose<3.1mmol/L.TheMDwas‐0.22
(95%CI‐0.47to0.03;I2=86%indicatingsubstantialheterogeneity;p=0.007).
Likewise,resultsshowedastatisticallysignificanteffectfortype1DM(MD‐
0.3695%CI‐0.53to‐0.19)andasmallereffectfortype2DM(MD‐0.1095%CI
‐0.19to‐0.01).TheGRADEqualityofevidenceforthisoutcomewasverylow(seeTa‐
ble11)whichmeansweareuncertainwhetherFreeStyleLibreleadstofewerevents
withglucose<3.1mmol/Lwithin7hours.
45 Clinical effectiveness and safety
TimewithGlucose>10.0mmol/L(h)
Onestudy(N=224)provideddataforthisoutcome.Therewasevidenceofnoeffecton
timespentwithglucose>10mmol/Linindividualswithtype2DM.TheMDwas0.00
(95%CI‐1.44to1.44)TheGRADEqualityofevidenceforthisoutcomewasverylow
(seeTable11)whichmeansweareuncertainwhetherFreeStyleLibreleadstoless
timewithglucose>10.0mmol/Lforindividualswithtype2DM.
TimewithGlucose>13.3mmol/L(h)
Timespentwithaglucoseconcentration>13.3mmol/Ldetectionfavouredtheinter‐
vention.TheoverallMDwas‐0.39(95%CI‐0.75to‐0.03;I20%indicatingnoheteroge‐
neity,p=0.03).Howeverresultsshowedstatisticallysignificanteffectonlyfortype1
DM(MD‐0.3995%CI‐0.77to‐0.01)butnotfordiabetestype2(MD‐0.4095%CI‐1.52
to0.72).TheGRADEqualityofevidenceforthisoutcomewaslow(seeTable11)which
meansFreeStyleLibremayslightlydecreasetimewithglucose>13.3mmol/Lforindi‐
vidualswithtype1and2DM.
46 Clinical effectiveness and safety
Table11.Summaryoffindingstable‐glycaemicmeasuresFreeStyle Libre compared to self-monitoring of blood glucose in insulin-treated diabetes individuals
Patient or population: individuals with diabetes type 1 and 2 insulin dependent Setting: home setting Intervention: FreeStyle Libre technology Comparison: SMBG Outcome: measured at 6 months (end of intervention)
Outcomes Anticipated absolute effects* (95% CI) Relative effect (95% CI)
№ of partici-pants
(studies)
Quality of the evidence (GRADE) Risk with SMBG Risk with FreeStyle
Libre technology
Time in range (hours) 3.9-10 mmol/L
The mean time with glu-cose range 3.9-10 mmol/L across control groups was 15 to 13 hours
The mean time (hours) with glucose range 3.9-10 mmol/L in the intervention group was 1 higher (0.32 higher to 1.68 higher)
- 462 (2 RCTs)
⨁⨁◯◯ LOW b
Hypoglycaemia
Time (hours): glucose < 3.9 mmol/L within 24h
The mean time with glu-cose < 3.9 mmol/L within 24h ranged across con-trol groups from 3 to 1hours
The mean time (hours) with glucose < 3.9 mmol/L within a period of 24h in the intervention group was 0.6 lower (0.88 lower to 0.32 lower)
-
462 (2 RCTs)
⨁◯◯◯ VERY LOW
b,c
Events (number): glu-cose < 3.9 mmol/L within 24h
The mean of glucose < 3.9 mmol/L within a pe-riod of 24h ranged across control groups from 0.5 to 2 events
The mean events (number) of glucose < 3.9 mmol/L within a period of 24h in the intervention group was 0.23 lower (0.35 lower to 0.1 lower)
- 462 (2 RCTs)
⨁◯◯◯ VERY LOW
b,d
Nocturnal Hypoglycae-mia Time (hours): glucose < 3.1 mmol/L during a pe-riod of 7h (23:00-06:00)
The mean nocturnal time with glucose < 3.1 mmol/L in the control groups range from 0.2 to 0.7 events
The mean nocturnal time with glucose < 3.1 mmol/L in the in-tervention group was 0.22 lower (0.46 lower to 0.03 higher)
- 462
(2 RCTs)
⨁◯◯◯
VERY LOW b,e
47 Clinical effectiveness and safety
FreeStyle Libre compared to self-monitoring of blood glucose in insulin-treated diabetes individuals
Patient or population: individuals with diabetes type 1 and 2 insulin dependent Setting: home setting Intervention: FreeStyle Libre technology Comparison: SMBG Outcome: measured at 6 months (end of intervention)
Outcomes Anticipated absolute effects* (95% CI) Relative effect (95% CI)
№ of partici-pants
(studies)
Quality of the evidence (GRADE) Risk with SMBG Risk with FreeStyle
Libre technology
Nocturnal Hypoglycae-mia Events (number): glu-cose < 3.1 mmol/L dur-ing a period of 7h (23:00-06:00)
The mean glucose < 3.1 mmol/L in the control groups ranged from 0.1 to 0.3 events
The mean glucose < 3.1 mmol/L in the in-tervention group was 0.09 lower (0.13 lower to 0.04 higher)
- 462 (2 RCTs)
⨁⨁◯◯ LOW b
Hypoglycaemia
Time (hours): glucose < 3.1 mmol/L within a pe-riod of 24h
The mean glucose < 3.1 mmol/L within a period of 24h ranged from 0.37 to 1.65 hours
The mean events (number) of glucose < 3.1 mmol/L within a period of 24h in the intervention group was 0.49 lower (1.15 lower to 016 higher)
- 462
(2 RCTs)
⨁◯◯◯ VERY LOW
b,f
Events (number) of glu-cose < 3.1 mmol/L within a period of 24h
The mean events (num-ber) of glucose < 3.1 mmol/L within a period of 24h ranged across con-trol groups from 0.24 to 0.92
The mean events (number) of glucose < 3.1 mmol/L within a period of 24h in the intervention group was 0.22 lower (0.47 lower to 0.03 higher)
462 (2 RCTs)
⨁◯◯◯ VERY LOW
b,g
Hyperglycemia Time (hours): glucose > 10.0 mmol/L
The mean time (hours) with glucose > 10.0 mmol/L in the control groups was 10
The mean time (hours) with glucose > 10.0 mmol/L in the intervention group was 0 (1.44 lower to 1.44 higher)
- 224 (1 RCT)
⨁◯◯◯ VERY LOW
b,h
Time (hours): glucose > 13.3 mmol/L
The mean time (hours) with glucose > 13.3 mmol/L ranged across control groups from 2 to 4
The mean time (hours) with glucose > 13.3 mmol/L in the intervention group was 0.39 lower (0.75 lower to 0.03 lower)
- 462 (2 RCTs)
⨁⨁◯◯ LOW b
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RCT: randomised control trial:
b. Issues related to unclear or high risk of selection and performance biases (downgraded twice) c. <3.9 mmol/L within 24hours/h: substantial heterogeneity (I2 84%) d. <3.9 mmol within 24hours /event: substantial heterogeneity (I2 64%) e. Nocturnal <3.1mmol (h): substantial heterogeneity (I2 85%) f. <3.1 mmol within 24hours/h: substantial heterogeneity (I2 89%) g. <3.1 mmol within 24hours/event: substantial heterogeneity (I2 86%) h. Imprecision: single study with small sample size (number of participants bellow 300 rule of thumb)
48 Clinical effectiveness and safety
Adverse Events
Non‐devicerelatedadverseeventsweremeasuredatbaselineand6months
endpointfortheinterventiongroup,andbaselineand4weeks(14daysat3
monthsandagainat6month)intheSMGBgroup.
Seriousadverseevents(non‐devicerelated)
Bothincludedstudiesprovideddataonseverenon‐devicerelatedadverseevents.Com‐
paredtoSMBG,FreeStyleLibrewasassociatedwithsimilarnumberofseriousadverse
events,RR0.74(95%CI0.19to2.85),I2=0%indicatingnoheterogeneity.
TheGRADEqualityofevidenceforthisoutcomewaslow(seeTable12)whichmeans
FreeStyleLibremayleadtolittleornodifferenceinseriousadverseeventsforindivid‐
ualswithType1and2DM.
Table12.Summaryoffindingstable‐adverseeventsFreeStyle Libre compared to self-monitoring of blood glucose in insulin-treated diabetes individuals
Patient or population: individuals with diabetes type 1 and 2 insulin dependent Setting: home setting Intervention: FreeStyle Libre technology Comparison: SMBG Outcome: measured at 6 months (end of intervention)
Outcomes Anticipated absolute effects* (95% CI)
Relative effect (95% CI)
№ of participants (studies)
Quality of the evi-dence
(GRADE)
Adverse Events Moderate to Severe
26 per 1000 19 per 1000 (5 to 73)
RR 0.74 (0.19 to 2.85
462 (2 RCTs)
⨁⨁◯◯ LOW b
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RCT: randomised control trial:
b. Issues related to unclear or high risk of selection and performance biases (downgraded twice) Devicerelatedadverseevents
Bolinder2016:“13adverseevents,reportedby10participantsintheintervention
group,wererelatedtowearingthesensor…therewere248sensorinsertion‐sitesigns
andsymptomsexperiencedby65participantsacrossbothgroups(26%ofpartici‐
pants).Signscanbesubdividedintothoseexpectedduetosensorinsertion:pain(38),
49 Clinical effectiveness and safety
bleeding(25),oedema(8),induration(5),andbruising(5),andthoseassociatedwith
sensorwear:erythema(85),itching(51),andrash(31).”page2260
Haak2017:“six(4.0%)interventionparticipantsreported9device‐relatedadverse
events(2severe,6moderate,and1mild).Theseweresensor‐adhesivereactions,pri‐
marilytreatedwithtopicalpreparations…therewere158anticipatedsensorinsertion
sitesymptomsobservedfor41(27.5%)interventionand9(12%)controlparticipants.
Thesesymptomswereprimarilyduetosensoradhesive(erythema,itching,andrash)
andresolvedwithoutmedicalinterventions.”
Sevencardiaceventswerereportedfor4(2.7%)interventionandthree(4.0%)control
participants(noneconsideredtoberelatedtostudyproceduresorthedevice).
Withdrawaldevicerelated
Bolinder2016:“Sevenparticipantswithdrewfromthestudyduetodevice‐related
eventsorrepetitiveoccurrencesofsensorinsertion‐relatedsymptoms”.
Haak2017:“Threeparticipants(1intervention,2controls)experiencedanadverse
eventleadingtowithdrawalsfromthestudy;nonewereassociatedwiththedevice”.
Inregardstoadverseevents,TheLancetpublishedcorrespondencebetweenBrahimi
(27)andBolinder(25)regardingconcernsaboutskinadverseeventswhichmightaf‐
fectindividualsusingtheFreeStyleLibredevice.Assomeinformationprovidedby
Bolinderwasnotaddressedinthesubmitter’spackageorthejournalpublications(i.e.,
useofadifferentbodysiteforthesensor,designchangesinthetechnology)weconsid‐
ereditimportanttopresentithere.
(Brahimi)“…theauthorsreported13cutaneousadverseeventsrelatedtotheuseof
FreeStyleLibre.Theseadverseeventsoccurredintenpatients,andwerecategorisedas
mild(threecases),moderate(fourcases),andsevere(sixcases).Twopatientswithse‐
vereskinadverseeventsweretreatedbydrugtherapyandtwowerediscontinued
fromthestudy.Allsevencaseswithmildormoderateskinadverseeventsrequired
drugtherapies,andthreeofthemwerediscontinuedfromthestudy.Itseemsreasona‐
blethatpatientswithskinadverseeventsmightdecidetocontinueordiscontinuetheir
participationinthetrial,butthemanagementoftheseadverseeventsduringthisstudy
remainsunclearandnotproportionaltotheseverityoftheadverseevents.Itwouldbe
helpfulforbothpractitionersandpatientsiftheauthorscandetermineclearlywhat
50 Clinical effectiveness and safety
typeofdevice‐relatedskinadverseeventsrequiretreatmentordevicediscontinuation,
orboth.”
Bolinder’sreply:
…thenumberandtypeofeventsreportedforFreeStyleLibreintheIMPACTtrialwere
similartothoseforothersystemsinwhichadeviceiswornontheskinforaperiodof
timeusingamedical‐gradeadhesive….Skinsymptomscanoccurwithhighskintem‐
peratureandhumidity,alongwithlongdurationofexposure,allofwhichmightbecon‐
tributingfactorstoadverseevents.Sensor‐wear‐relatedsymptomswererecordedas
adverseeventsintheIMPACTtrialiftheeffectsweresevereandlastedformorethan7
days,orifthepatientrequiredprescriptionmedicationfortheeventtoresolve.Ad‐
verseeventseveritieswererecordedonthebasisofahealth‐careprofessional’sassess‐
mentofmild,moderate,orsevereevents.
Accordingtothestudyprotocol,individualswithknownsensitivitytomedical‐grade
adhesiveswereexcludedfromparticipation.However,wereasonablyexpectedthata
fewparticipantsmighthavebeenunawareoftheirsensitivity...Forparticipantswith
adverseeventsinvolvingskinsymptomsduringthistrial,symptoms(includingsevere)
wereresolvedbyuseofbarrierproducts(e.g.,Cavilonspray)ordrugtherapy(e.g.,zinc
ointment,Fenistilgel,orhydrocortisonecream)asprescribed,orsimplybyrelocating
thedevicetoanotherareaoftheskinsuchthattheeffectsweremaintainedatatolera‐
ble,backgroundlevel.Inothercases,althoughtheadverseeventsweregenerallymild
ormoderate,thelongevityofthesymptoms,despiteuseoftreatment,contributedto
theparticipant'sdecisiontowithdrawfromthetrial.Noneoftheparticipantswithdrew
becauseofhealth‐careprofessionaladvicetostopwearingthesensor.
SincecompletionoftheIMPACTtrial,minordesignchangeshavebeenmadetoFree‐
StyleLibre.Thesechangesareexpectedtoimprovebreathabilityoftheskinthatisin
contactwiththesensorandtofacilitatetheexclusionofmoisturebetweenthesensor–
skininterface.
51 Cost-effectiveness analysis
Cost‐effectivenessanalysis
Methods for evaluating submitted cost-effectiveness models
Cost-effectiveness analysis
Theprimaryobjectivesofhealtheconomicmodellingaretoevaluatetheincremental
cost‐effectivenessofthespecifiedhealthintervention(s)comparedtostandardtreat‐
ment,usingthebestavailableevidence,andtoassessthemostimportantsourcesof
uncertaintysurroundingtheresultsandtherobustnessoftheresults.Inordertomake
comparisonsacrossdifferenttypesoftreatmentsandmultiplehealthoutcomes,eco‐
nomicmodelstypicallymeasuretreatmentbenefitsintermsofquality‐adjustedlife
years(QALYs),ameasuredesignedtocapturetheutilityofbothlifeextensionand
healthimprovement.QALYs,bydefinition,takeonavalueof1forperfecthealthand0
atdeath(28).Theoutputofacost‐effectivenessmodelisexpressedasanincremental
cost‐effectivenessratio(ICER),whichcanbethoughtofastheextracostofobtainingan
extralife‐yearinperfecthealth.TheICERisdefinedas
Evaluating cost-effectiveness models
Thereisnosinglecorrectwaytobuildaneconomicmodeltoestimatethecost‐effec‐
tivenessofaspecifichealthinitiative.Modellingrequiresconsultingwithclinicalex‐
pertstogainanunderstandingofnormaldiseaseprogression,andtodetermine,based
ontheresearchquestion,therelevanttreatmentpopulation,relevantcomparator,and
importanthealthoutcomesandadverseeventsconnectedtotreatment.Thisinforms
thebasicmodelstructure,andalsodetermineswhichclinicaleffectdataismostim‐
portanttoretrieveinthesystematicliteraturesearch.Oncethemodelstructureisin
place,systematicsearchesandevidencegradingareusedtoprovidethemostreliable
riskinformationforthemodel,butmustalsotocollectalloftherelevantcostandqual‐
ityoflifedatathatisneededforcost‐effectivenesscalculations.
52 Cost-effectiveness analysis
Amodelisrarelymeanttocaptureeverypotentialdetailofthetreatmentlandscape;
ratherthegoalistoincludeenoughdetailtoprovidearealisticviewofthemostsignifi‐
cantpathwaysindiseaseprogression,giventheresearchquestion(s)oneistryingto
answer.Evaluatinganygivenmodelisprimarilyaboutdeterminingwhetherthe
choicesmadebythesubmitterregardingmodelstructureandtreatmentcomparator
arereasonablegiventheresearchquestion;whetherbaselineepidemiologicaldatare‐
flectthepopulationinwhichtheanalysisisbeingperformed;whethertheclinicaleffect
datausedinthemodelareofadequatequality;whetherresourceuseandcostsreflect
theconditionsofthehealthcaresysteminquestion;whethertherehasbeensufficient
sensitivityandscenarioanalysistodeterminethedegreeandsourceofuncertaintyin
themodelresults;andwhetherthemodeldisplaysexternalandinternalvalidity.
Checklistsareavailabletohelpresearcherssystematicallyexaminetheseissues(28).
Weproceedbyfirstdescribingthehealtheconomicmodelusedinthesubmissionand
theresultsgeneratedbythemodel.Wethenprovideourevaluationofthemodel,fo‐
cusingonthefollowingissues:modelstructure,choiceofmodelparameters,useofap‐
propriatesensitivityand/orscenarioanalysistoexaminetheextentofuncertaintyin
modelresults,andrelevanceofthemodelfortheNorwegiancontext(28).
Publishedcost‐effectivenessevaluationsidentifiedbythesubmitter
Thesubmitteridentifiedandprovidedtwopublishedcost‐effectivenessevaluationsof
flashglucosemonitoringsystems.OneeconomicanalysisexaminedEuropeanand
Australianindividualsreceivingintensiveinsulintreatmentfortype1DM(29).The
secondeconomicanalysisfocusedonEuropeanindividualsreceivingintensiveinsulin
treatmentfortype2DM(30).Bothstudieswerefundedbythesubmitterandwere
usedtheIMSCoreDiabetesModel(Table13).TheNIPHsearchedforotherpublished
economicevaluationsofFreeStyleLibre,withouthfindinganyrelevantones.
53 Cost-effectiveness analysis
Table13.FreeStyleLibrecost‐effectivenessevaluationsfortype1and2DM
CEA:Cost‐effectivenessanalysis,CDM:Corediabetesmodel,SEK:Swedishcrones,ICER:incrementalcosteffectivenessratio,QALY:qualityadjustedlifeyear,SMBG:self‐monitoringbloodglucose
Descriptionofthepublishedcost‐effectivenessevaluationsidentifiedbythe
submitter
Biliretal.(29)estimatedthecost‐effectivenessofFreeStyleLibrecomparedtoSMBG
throughtheIMSCoreDiabetesModel(IMSCDM)forintensiveinsulin‐treatedType1
DM.TheIMSCDMcombinesMarkovmodelstructureswithaMonteCarlosimulation.
HbA1cprogressionwasbasedondatafromtheDiabetesControlandComplications
Trialstudy(31),whileotherphysiologicalparametersprogressionweretakenfromthe
FraminghamHeartStudy(32).Swedenwasthecorecase,withadditionalresultsfor
Germany,Italy,France,theNetherlandsandAustralia.Forthecorecase,cost‐
effectivenesswasmodelledfromasocietalperspectivewitha50‐yeartimehorizon.
TheestimatedICERforFreeStyleLibrecomparedtobloodglucosemonitoringwas
240,826Swedishcrones(SEK)/QALY.
Study Bilir et al (29) Li et al (30)
Model Analysis CEA CEA
Population Individuals >18 years with well-controlled
Type 1 diabetes, HbA1c of ≤7.5% (58
mmol/mol) treated by multiple daily insulin
injections or continuous subcutaneous
insulin infusion for at least 6 months (25).
Glucose levels were self tested by
individuals at least 10 times per week.
Individual characteristics in the analyses
reflect the IMPACT trial population and
estimates from the published literature were
used for CDM inputs unavailable from the
IMPACT study.
Individuals >18 years with poorly controlled
Type 2 diabetes, HbA1c of ≥7.5% (58
mmol/mol) treated by multiple daily insulin
injections or continuous subcutaneous
insulin infusion for at least 6 months
(REPLACE trial). SMBG was done by
individuals at least 10 times per week.
Individuals’ characteristics in the analyses
reflect the REPLACE trial population and
estimates from the published literature were
used for inputs unavailable from the
REPLACE study.
Intervention FreeStyle Libre FreeStyle Libre
Comparison SMBG SMBG
Incremetal QALY 0.80
0.56
Incremental costs SEK192,973
SEK144,360
ICER/QALY SEK240,826/QALY
SEK258,108/QALY
54 Cost-effectiveness analysis
Lietal.(30)estimatedthecost‐effectivenessofFreeStyleLibrecomparedtoSMBG
throughtheIMSCDMforintensiveinsulin‐treatedType2DMusingthesameIMSCDM
modelasBiliretal.TheyalsousedSwedenasthecorecase,withadditionalresultsfor
Germany,Italy,FranceandtheNetherlands.TheSwedishNationalDiabetesRegister
riskequationwasusedforHbA1cvalueprediction,whiletheCDMdefaultriskequation
(basedonUnitedKingdomProspectiveDiabetesStudy(33))wasusedforother
countries.Otherphysiologicalparametersprogresiondataweretakenfromthe
FraminghamHeartStudy(32).Thecorecase,cost‐effectivenesswasmodelledfroma
societalperspectivewitha40‐yeartimehorizon.TheICERforFreeStyleLibre
comparedtobloodglucosemonitoringwas258,108SEK/QALY.
Scenarioanalysesperformedfortheothercountries,inbothstudies,showedsmiliarresultsastheSwedishcorecase.
Populationandcomparatorinthesubmittedreport
Thecost‐effectivenessanalysispopulationareindividualswithtype1andtype2DM.
Themaincharacteristicsofindividualswithtype1DMareage≥18years,HbA1clower
than7.5%,andperformedSMBGatleast3timeseachdayonaverage(25).Themain
characteristicsofindividualswithtype2DMareage≥18years,HbA1cbetween7.5%
and12%,andSMBGperformedatleast10timesaweekonaverage(26)(Appendix5).
ThesubmitterpresentedtheanalysisofFreeStyleLibrecomparedtoSMBG.Thelatter
isthecurrentreferencetreatmentforinsulindependentindividualswithtype1and2
DMinNorway.
Typeofanalysisanddecisionmodelofthesubmittedreport
ThesubmittedreportusedtheIMSCDM,whichwasdesignedtoassesslifetimehealth
outcomesandeconomicconsequencesofvariousmeasuresforDM.Themodeldeter‐
minestheICERperQALY.Theanalysiswasconductedfromtheperspectivesofboth
thehealthcareservicesandsocietymorebroadly,anduseda40‐yeartimehorizon.
Costsandclinicaloutcomesareextrapolatedinthemodelbeyondthestudyfollow‐up
periodof6months.Thediscountrateintheanalysiswassetto4%forbothcostsand
QALYs.Themodelincludedtwoarms:theFreeStyleLibreandSMBGarm.Theanalysis
assumedthattheindividualswithdiabetesusedeitheronlySMBGorFreeStyleLibre
continuouslythroughoutthelifetime.ThemodelisMarkov‐basedwithannualcycles
55 Cost-effectiveness analysis
meaningthatanindividual’shealthstatewithcostandeffectimplicationsisevaluated
atannualintervals.
Accordingtothesubmitterthemodelquantifiesthedevelopmentofcomplications,life
expectancy,quality‐adjustedlife‐yearsandtotalcostsforthecohortinthestudyover
the40‐yearprojectiontime.Baselinecharacteristicsforindividualswithtype1DMare
fromtheIMPACTstudyandreflectthemeanbaselinecharacteristicsofthestudy’s
treatmentandcontrolgroups:ageof44years,HbA1cof6.78%,andmeandiabetesdu‐
rationof22years(25).Similarly,baselinecharacteristicsfortype2DMcohortarefrom
theREPLACEstudy:ageof59,HbA1cof8.70%,andmeandurationofdiabetesof17.5
years(26)(seeTable6).
Themodelstructurecomprises17sub‐models,independentofeachother,thatsimu‐
latethecomplicationsofdiabetes(Figure7).Eachsub‐modelisaMarkovmodelusing
MonteCarlosimulationstopredictoutcomes.Transitionprobabilitiesrepresentedby
distributionspredictoccurrenceofcomplicationduringayear,anddependontimeand
thecurrenthealthstatus.Themodelusesdifferenttransitionprobabilitiesandproce‐
duresfortype1andtype2DM.Fortype1DMthemostimportantdatasourcesarethe
diabetescontrolandcomplicationstrialandtheFraminghamHeartStudystudies(31).
Fortype2DMthemostimportantdatasourceistheUnitedKingdomProspectiveDia‐
betesStudy(33).Becauseeightofthe17diabeticcomplicationsmayresultinafatal
outcome,themodelalsocalculatesbackgroundmortality.
AccordingtothesubmitterthemodelusestheC++(Microsoft®VisualStudio2005)
programtoformthebasisofthecalculationsrequiredtoruneachsimulation(31).
56 Cost-effectiveness analysis
Figure 7. Flow diagram of the IMS Core Diabetes Model. PVD: peripheral vascular disease, MI: myocardial infarction,
CVD: cardiovascular disease, CHF: congestive heart failure, ARB: angiotensin receptor blocker, ACE: angiotensin-
converting enzyme (34).
Generalcommentsonthesubmittedhealtheconomicanalysis
Themodelutilisedbythesubmitterlackstransparency,thisshallbeunderstoodasthe
difficultyfortheNPIHtogainafirmunderstandingofthefactorsthatdeterminehowa
singlepatientprogressthroughthemodel,whichassumptionsaremadeandwhich
parametersaffectthemodel´soutcomes.Specifically:
wedidnothaveinformationaboutseveralkeyequationsandassumed
relationshipsinthemodel,
welackedaccesstoamicrosimulationmodulewhereitshouldbe
possibletorunsensitivityanalyses(35),
welackedaccesstotheassumeddistributionofvariables,whichiscrucial
fortheassessmentofaMonteCarloprobabilisticmodel.
Patientleveldataaremissingfromthemodelthatwehaveaccessto(35).
57 Cost-effectiveness analysis
Theproblemwithlackoftransparencyisreinforcedbytheunexpectedbehaviorofthe
model.Themodelgeneratedresultsthatwerehardforustoexplaingiventhe
informationprovidedbythesubmitteraboutthemodel.Forexample,thesubmitter
statedinameetingthatHbA1cwasthemaindriverfortheanalysislinkedtotype1and
2DM.WefoundthatchangesinHbA1cmadelittlechangeintheICERs.
Itwasalsodifficulttoassesstheinternalvalidityofthemodel.Becauseofourlackof
fullaccesstothemodel,itwasimposibletoperformafullassessmentortomodify
underlyingassumptionsandparametersinordertoindependentlyassesstheimpact
onreportedresults.Ouraccesswaslimitedtoaweb‐basedmodel,whichonlyallowed
ustoexaminethemodel'sinputdata.
ThesubmitterarguesthattheIMSCoreDiabetesModelhasbeenusedinmanypeer
reviewedpublications,andhasalsobeenvalidatedinanarticlepublishedinahigh
qualityjournal(34;36).Forvalidationpurposes,McEwanetalusedthemodelto
predictclinicaloutcomesin112clinicaldiabetesstudiesbasedonbaseline
characteristicsfromthestudies,andthencomparedthemodel’sresultswiththe“gold
standard”clinicaltrialoutcomes.TheresultingR2valuesof0.90and0.88fortype1and
2DMrespectively,indicatethatthemodelwasabletoexplainalargeportionofthe
variationinresultsofthe112clinicalstudies.TheythereforeconcludethattheIMS
CoreDiabetesModelisacredibletoolforpredictingtheabsolutenumberofclinical
episodesinthepopulationsfromtheDiabetesControlandComplicationsTrialandthe
UnitedKingdomProspectiveDiabetesStudy(33).Ourclinicalexpertsfoundthe
assumedpatientpopulationofthesubmissiontobereasonableforNorwegian
conditions.
Althoughthesubmittedanalysisexaminedbothahealthcareperspectiveandasocietal
perspective,ourassessmentfocusedonthehealthcareperspective.Ahealthcare
perspectiveisappropriatewhendecision‐makingoccurswithafixedbudget,asisthe
casefortheNorwegianRegionalHealthAuthorities.
Clinicalandepidemiologicaldata
ThesubmittedIMSCoreDiabetesModelusesseparatetransitionprobabilitiesand
strategiesofmanagementfortype1andtype2DM.Modelparametersarebasedon
datafromseveralpublishedclinicalandepidemiologicalstudiesfromdifferent
countries.
58 Cost-effectiveness analysis
Themodelisdescribedtocapturethenaturalcourseofdiabetes,associatedwiththe
relationshipbetweenriskfactorsandspecificevents,mortalityfromcomplicationsand
long‐termmortality.Themodelincludesincidencedataforthefollowingadverse
eventsamongpatientswithtype1andtype2DM:fatalcardiovascularevents(stroke,
myocardialinfarction,heartfailureandangina),non‐fatalcardiovascularevents,
microvascularcomplications,footulcerandamputation,depressionandothermildor
severeadverseevents(i.e.,hypoglycemia,ketoacidosis,lacticacidosis).
Inthedefaultscenariotheamountofinsulinconsumedorrequiredisnotintensified
overtime.Thisisbasedon30monthsofdosingdatafromBuseandcolleagues(37).
Efficacy
TheefficacydatainthesubmittedmodelistakendirectlyfromtheIMPACTandRE‐
PLACEclinicalstudies(bothevaluatedovera6‐monthperiod).Efficacydatabasedon
intention‐to‐treatanalysisfromtheclinicalstudiesareassumedtoreflecttheinterven‐
tioneffectinthemodel.Thiseffectisbelievedtopersistthroughoutthestudy’stime
horizon.
TheclinicaleffectivenessofFreeStyleLibreforindividualswithtype1DMwasmeas‐
uredasthedifferencebetweenthegroupsintimeperdayspentinhypoglycaemia(glu‐
cose<70mg/dL[3.9mmol/L]),andinthemodelappliedbyadjustingthebaselineval‐
ues.IntheIMPACTtrial,theprimaryendpointoftimespentinhypoglycaemiawassig‐
nificantlyreducedintheinterventionarmcomparedtothecontrolgroup.Atbaseline,
themeantimespentinhypoglycaemiawas3.38hoursperdayintheFreeStyleLibre
armand3.44hoursinthebloodglucosearm,representinga2%reduction.Afterad‐
justmentsforbaselinehypoglycaemia,theadjustedmeandifferencebetweenthe
groupswas−1.24hoursperday,correspondingtoa38%reductionoftimewithhypo‐
glycaemiawithFreeStyleLibrecomparedwithSMBG(25).
Forindividualswithtype1DMintheSMBGarm,therateformildhypoglycaemic
eventsis6,760per100person‐year(27%oftheseeventswerenocturnalhypoglyce‐
micevents).Forindividualswithtype1DMintheFreeStyleLibrearm,therateformild
hypoglycaemiceventswascalculatedbasedon25.5%and33.2%reductioninthefre‐
quencyofdailyandnocturnalhypoglycemicevents,respectively(glucose<70m/dL).
Therateofwhatturnedouttobe4,897per100person‐years,(25%oftheseevents
werenocturnalhypoglycemicevents)(25).
59 Cost-effectiveness analysis
AccordingtotheUnitedKingdomHypoglycemiaStudyGroup(2007)(38)theratesfor
severehypoglycemiaarethesameinbotharms.Theanalysisisbasedonthis.
Individualswithtype1DMinboththeFreeStyleLibreandSMBGarm,hadthesame
meanchangeinHbA1c(0.12%)atbaseline.ThiswaspresentedintheIMSCoreDiabe‐
tesModel.
TheclinicaleffectivenessofFreeStyleLibreforindividualswithtype2DMwasmeas‐
uredasthedifferencebetweenthechangesinHbA1c.Thestudywaspoweredtodetect
adifferenceinthechangeinHbA1cat6monthsof0.35%.Theprimaryendpointofthe
REPLACEstudy,reductioninHbA1c,wassimilarforbotharms.Theadjustedmean
changeinHbA1cfrombaselinewas−0.29%inFreeStyleLibrecomparedto−0.31%in
theSMBGgroupwithnosignificantdifferencebetweengroups(26).
Informationaboutthenumberofmildhypoglycaemiceventsforindividualswithtype
2DMintheSMBGarmwerefoundinarecentlypublishedmeta‐analysisstudyfor
SMBG.Forindividualswithtype2DMusingFreeStyleLibre,thenumberofmildhypo‐
glycemiceventswerecalculatedbasedontherelativedifferencebetweenthetreat‐
mentarms.Therateofmildhypoglycemiceventswas27.7%lowerintheFreeStyleLi‐
brearmcomparedtotheSMBGarmpresentedintheonline‐basedmodelforindividu‐
alswithtype2DM(26).Further,nodifferencewasassumedinseverehypoglycemic
eventsbetweenthearms.Theseratesareassumedtobeconstantduringthe40‐year
timehorizon.
Thesubmitterhasalsoincorporatedanadditionalutilityrateof0.03intheIMSCore
DiabetesModelfortheuseofFreeStyleLibrecomparedwithSMBGinalltheirscenar‐
ios,toaccountfortheassumedimprovementinuser‐friendliness.Theutilityvalueof
0.03(andotherutilityvalues)forFreeStyleLibreisbasedonaTimeTradeOffstudy,
performedbyEvideraandsponsoredbyAbbottDiabetesCare.TheEQ‐5Dhaslimited
sensitivitytoutilitybenefitsassociatedwithglucosemonitoringdevices,andisthere‐
forenotincludedintheREPLACEorIMPACTstudies(25;26).
Costs
Thesubmitteridentifiedresourceuseandcostdatabysearchinginpublished
Norwegiancoststudiesandadministrativedatabases.Allcostsarefromtheyear2016
andmeasuredinNorwegiankroner(NOK)andtheconsumerpriceindex(CPI)isused
60 Cost-effectiveness analysis
toadjustcostsfromdifferentyears(39).Thesubmitteddocumentationpackage
containsdirectcostsrelatedtothehealthcareperspective,andindirectcostsrelatedto
asocialperspective.
ThecostofFreeStyleLibreincludesthecostoftheintervention,thecostoftreatment
forcomplications,thecostrelatedtosideeffectsandothertreatmentcosts.
Cost of the intervention
ThesubmittercalculatedthetotalyearlycostsrelatedtoFreeStyleLibreandstandard
SMBGforindividualswithtype1and2DM(seeTable14).Thecalculatedcostof
FreeStyleLibreislistedintable14.
Table14:CalculatedcostofFreeStyleLibre
Item Cost (NOK) Unit Source (REF)
Insulin 0.34 IU/ml (40)
Metformin 0.38 500 mg x 3 per
day
(40)
FreeStyle Libre 599 (excl. VAT) Per sensor ((41), read June, 2017)
Test strips 4.82 Per strip (42)
Lancets 0.53 Per lancet (43)
Additional GP visit 284
(x 2 for FreeStyle Libre year one)
Per consultation
(44)
GP: general practitioner, mg: miligram; SLV: The norwegian medicines agency; IU/ml: international
units per milimiter; VAT: Valute-Added Tax
TheunitpriceoftheFreeStyleLibresensor(NOK599)istakenfromthewebsiteofFreeStyleLibre(41).Thepricestatedinthesubmissionwaslower,butclaimedbythesumbittertobeconfidential.
Diabetestype2<65yearsold:
Forresourceusecalculation,thefollowingassumptionsobservedintheREPLACE
study(26)wereused.General:anindividualwithtype2DMusesmetformininaddition
toinsulin.
FreeStyleLibre:Thesubmitterassumethattheindividualneeds109.5SMBGstripsper
year,85.2insulinunitsperdayand0.65lancetsperday.Further,thesubmitter
suggeststhattheindividualneedstwoGPconsultationsthefirstyearandone
consultationinsubsequentyears.ThecostforFreeStyleLibreuserswouldbeabout
NOK24,954thefirstyear,andNOK24,670thesecondyear.Accordingtothewebsiteof
61 Cost-effectiveness analysis
FreeStyleLibreasensorworksuptofourteendays(41).Ourclinicalexperts
consideredthatanindividualneedsbetween26to29sensorsperyear.
SMGB:ThesubmitterassumesthattheindividualneedsthreeSMBGstripsperday,87.8
insulinunitsperdayand1.26lancetsperday.Further,thesubmittersuggeststhatthe
individualneedsoneGPconsultationperyear.TheannualcostforSMBGuserswould
beaboutNOK17,116.
Diabetestype1:
ThefollowingassumptionslaybehindthedataintheIMPACTstudy(25).
FreeStyleLibre:ContinuousglucosemonitoringsystemsinNorwayaremainlyusedby
thoseindividualswithimpairedawerenessofhypoglycaemia.
Thesubmitterassumesthattheindividualneeds182.5SMBGstripsperyear,45.8
insulinunitsperdayand267.4lancetsperyear(25).Accordingtothesubmitter’s
websiteFreeStyleLibresensorworksuptofourteendays(41).Ourclinicalexperts
consideredthatanindividualneedsbetween26to29sensorsperyear.
Further,thesubmittersuggeststhatoneindividualneedstwoGPconsultationsinthe
firstyearandoneconsultationinsubsequentyears.ThetotalcostforFreeStyleLibre
userswouldthenbeaboutNOK23,446inthefirstyearandNOK23,162inthesecond
andsubsequentyears.
SMBG: Thesubmitterassumesthatoneindividualneeds1,971SMBGstripsperyear,
34.8insulinunitsperdayand657.6lancetsperyear(25).Further,thesubmitter
suggeststhatoneindividualneedsoneGPconsultationeachyear.Theannualcostfor
SMBGuserswouldbeaboutNOK14,904.
Cost of complications
Thesubmitterobtainedcostsrelatedtomyocardialinfarction,angina,strokeand
peripheralvasculardiseasebasedonNorwegiandiagnoses‐relatedgroups(DRGs)(45).
Thefirstyearcostsrelatedtomyocardialinfarction(NOK53,311),angina(NOK
21,840),heartfailure(NOK47,299)andperipheralvasculardisease(NOK39,259)are
basedonDRGs(45).ThecostrelatedtostrokeinthefirstyearisaboutNOK91,266and
basedonseveralpublishedsources(45‐47).
ThesubmitteradoptedtheassumptionofthereimbursementreportforPradaxathat
thecostsrelatedtomyocardialinfarctionwilldeclineby69%fromthefirstyeartothe
secondyearandbeyond(45;48).AccordingtothereimbursementreportbyLevemir
(46)andadjustmentsbyCPIfrom2007to2016oncostdata(39),thecostrelatedto
62 Cost-effectiveness analysis
strokeandperipheralvasculardiseasewillbothdeclinetoNOK9,320inyeartwoand
beyond.Deathduetostrokewithin30daysisgiventhesamecostinthefirstyear,but
regardlessofthecostsrelatedtorehabilitation.Thecostrelatedtoanginainyeartwo
andbeyondareNOK415(40;44;45).Thesubmitterdoesnotmentionanysourcefor
thecostsassignedtoheartfailureinyeartwoandbeyond(NOK2,443).
Thecostsofdialysisinyearone(NOK792,887)(49)andthecostofperitonealdialysis
inyearone(NOK629,085)arealsousedinyeartwoandbeyond,minusthecosts
relatedtotrainingofhealthprofessionals.DRG,immunosuppressivedrugsand
corticosteroids,accordingtotheindicationofthedrugmycophenolateareusedto
estimatethetransplantationcosts(NOK539,558)(40;45).
Cost of the side effects
Inthesubmittedmodel,mildhypoglycaemiceventsareassumedtohavenocost.
SeverehypoglycaemiceventsaregivenacostofNOK31,583becausethesubmitter
assumethatindividualshavingseverehypoglycaemianeedhelpfromathirdparty.
Helpfromathirdpartyalsoappliestoindividualsaffectedbyketoacidosis,whichthe
submitterconsiderstocostNOK48,022.EdemaisgivenacostofNOK3,787(45).
Other treatment costs
TheinputcostsrelatedtomedicationswerederivedfromtheNorwegianMedicines
Agency(40)andthecostsrelatedtoscreeningandcataractsurgerywereobtained
fromtheNorwegianMedicalAssociationandUnilabs(44;50).Thecostrelatedtolaser
treatment,treatmentforblindness,neuropathy,amputation(eventbased),healed
woundandinfectedwoundweretakenfromtheNorwegianDirectorateofHealth(45).
Thecostrelatedtogangrenetreatmentandamputationprosthesiswereobtainedfrom
thestudybyGhatnekaretal.(51).
Healthrelatedqualityoflife(HRQL)
HRQLutilityvalues,basedonscoresfromthegenericEuroQolinstrument(EQ‐5D),
wereavailableforallhealthstates.MostoftheutilityvaluesweretakenfromEuropean
studies.Thehealthstate“Diabetestype2,withnocomplications”wasgivenautility
valueof0.785(52)andthehealthstate“Diabetestype1,withnocomplications”was
givenautilityvalueof0.9.Inthesubmittedmodel,theHRQLchangeovertime,depend‐
ingondisabilitiesrelatedtocomplicationsoreventsthatmayoccurforanindividual
withdiabetes.AnindividualstartswiththesamebaselineHRQLutilityvalueinboth
63 Cost-effectiveness analysis
studygroups(FreeStyleLibreandSMBG),andexperiencedisutilityassociatedwiththe
variousminorandmajordiabetesrelatedevents.
Theutilityvaluesformildhypoglycaemiaandthefrequencyofhypoglycaemiaare
basedoncalculationsfromLauridsenetal(53).Thefirstindividualexperiencewithhy‐
poglycaemiawillbeperceivedassevere.Thedisutilitypereventdecreaseastheindi‐
vidualexperienceshypoglycaemicepisodesmoreoftenandadapttothesituation.Simi‐
larly,theaveragedisutilitypereventdependsontherateofhypoglycaemia.Takingthis
intoaccount,thedisutilityperhypoglycaemiceventwasestimatedineachtreatment
arm.
NIPHcommentsonsubmitterparametersandinputdata
Thereisgreatuncertaintyconnectedtotheefficacydata.Inputdataforthemodelis
takenfromseveralstudies,fromdifferentcountrieswithvariousstudypopulations.
Thus,thestudypopulationsmaynotbecomparableandmaynotrepresentthecurrent
Norwegianpopulation.Accordingtoourclinicalexperts,whenusingolddataitshould
bepointedoutasapossiblesourceofbiasedresults.Historicdataoftenoverestimate
treatmenteffects.Alotofthedatausedforinputinthehealtheconomicmodelareold.
Usingdatafromthe50s,60s,70s’(32)andmostly80s’(33)would,forinstance,greatly
overestimatecardiovasculardiseaseinindividualswithDM,andhencethepotentialfor
savingsbydecreasingglucoselevels.Further,atthattimeprimarypreventionwith
lipidloweringdrugswasnotinventedandthebordersforinitiationofantihypertensive
treatmentwerehigher.
Thesubmittedclinicalstudiesextrapolatetheeffectof6monthsovera40‐yearperiod
(25;26).A6‐monthperiodwaschosen,becauseitislongenoughtodetectdifferences
indurationspentinhypoglicaemiathatmayresultfromparticipantsandhealthcare
professionalsrespondingtothemorecomprehensiveglucosedataprovidedby
FreeStyleLibre(25;26).Thelongtermeffectsoftreatmentofhypoglicaemia,andthe
longtermeffectsofhypoglicaemiaincidence,arelessknown,which
increasetheuncertaintyregardingtheefficacydatainthemodel.Patientfatiguein
wearingFreeStyleLibreisnottakenintoaccount.Neitheristhefactthatinitialstudies,
inparticularlythoseinvolvingdedicatedindividuals,tendtooverestimatetreatment
effectscomparedtowhenthetreatmentisusedinageneraldiseasepopulation(clinical
expert).
64 Cost-effectiveness analysis
Theclinicalinputdatausedforcost‐effectivenesscalculationspresentedinthemodel
deviatesubstantiallyfromtheclinicalinputdatapresentedinthesubmitted
documentationfileforbothtype1andtype2DM.Forindividualswithdiabetestype1,
theonline‐basedmodelusedclinicalinputdatafromtheREPLACEstudy(thisstudy
includesindividualswithtype2DM,under65years)inonesimulation,andclinical
inputdatarelatedtoaSwedishdatacollectioninanothersimulation.Theseconstitute
considerabledeviationsregardingclinicalrisksandprobabilitydatafromtheclinical
inputsfoundinIMPACTandREPLACE.ThemodeldidnothaveanyHbA1cadjustments
relatedtotype1DMforthecontrolgroup,butonlyfortheinterventiongroup.Inthe
submittedmodel,thechangeinbaselineHbA1cweredifferentbetweenthe
interventiongroupandthecontrolgroup.Accordingtotheliterature,thechangein
baselinehbA1cshouldbethesame.
Althoughtheraterelatedtomildhypoglycemiceventsweresomewhatlowerin
FreeStyleLibrearmcomparedtotheSMBGarmpresentedintheonline‐basedmodel
forindividualswithtype2DM,theseverehypoglycemiceventsweresimilarineach
arm.Thecostrelatedtoseverehypoglycemiceventsweremuchhigherthanmild
hypoglycemicevents.Thesubmitterassumesthateveryseverehypoglycemicevent
needsathirdpartyintervention(involvingahospital/doctor)andthengeneratesa
cost.However,oneofourclinicalexpertsstatesthatseveralpatientsmayhave
episodesofseverehypoglycemiathattheyhandlethemselveswithoutintervention
fromathirdparty.Andwhenathirdpartyisneededthatwouldmostoftenberelatives
athome.
ThesubmitterconsideredadifferenceinchangeHbA1cbetweenFreeStyleLibreand
SMBG.However,thestudiesfromBolinderandHaak(25;26)shownostatisticaly
significantchangeinHbA1cbetweeninterventionandcontrol.Thelackofstatistical
significanceisnotanerrorperse,butintroducesuncertaintythatmeritsfurther
considerationinone‐waysensitivityanalysis.Thiswasnotperformedbythesubmitter
andwasnotpossibletoexplorebythereviewersbecauseoftherigidityoftheIMS
onlinemodel.
Regardingthecostandresourceuse,ourclinicalexpertsconsiderthattheinsulincon‐
sumptionshouldbethesameifFreestyleLibreisused,giventhatdiabetesiscorrectly
treated.Inchildrenandyouthunder18yearsoldtheconsumptionofstripswould,ac‐
cordingtoourclinicalexperts,behigherthan5.4perdaywhichwasassumedbythe
submitter.Intheiropinion,severalchildrenusemorethan10stripsaday.
65 Cost-effectiveness analysis
Cost‐effectivenessresults
Thesubmitterprovidedseparatecost‐effectivenessresultsfortype1andtype2DM.
Thesubmittersresultsfortype1DMwerebasedonanequalchangeinHbA1cinboth
arms(0.12%),howeverresultsaredrivenbytheassumptionthatindividualsinthe
FreeStyleLibrearmhadautilitygainof0.03.Thesubmitter’sresultconsequentlyshow
thattheuseofFreeStyleLibregivesahigherQALYandalowercost,resultingina
dominantICER.
Table15.Cost‐effectivenessresults(type1DM)ofFreeStyleLibreversusSMBGaccordingtosubmitter’smodel
Population Incremental Cost (NOK) Incremental Effect (QALYs)
ICER
Type 1 DM -1 ,225 ,067 1.17 dominant NOK:NorwegianKroner;QALYs:qualityadjustedlifeyear;ICER:incrementalcosteffectivenessratioThesubmitters’resultsfortype2DMweregivenintwoscenarios,oneforall
individualsintheREPLACEstudyandoneforindividualsagedunder65years(Table
16).Thesubmitterassumedutilitygainwas0.03forFreeStyleLibreuse.Accordingto
thesubmitter,theuseofFreeStyleLibregivesahigherQALYandhighercosts,which
resultsinanICERofNOK235,673perQALYgained.Accordingtothesubmitter,the
scenariowithindividualsunder65years,theresultingICERisNOK243,434.
Table16.Cost‐effectivenessresults(type2DM)ofFreeStyleLibreversusSMBGaccordingtosubmitter’smodel
Population Incremental Cost (NOK) Incremental Effect (QALYs)
ICER (NOK)
Type 2 DM 88731 0.38 235673
Type 2 DM < 65 years 103119 0.42 243434 NOK:NorwegianKroner;QALYs:qualityadjustedlifeyear;ICER:incrementalcosteffectivenessratio
NIPH comments in the results
Wewerenotabletomakeourownresultsoradjustsubmittersresults,i.e.wewerenot
abletomakeourownincrementalcost‐effectivenessratios(ICERs).Thiswasbecause
thesubmitter’smodelwasnotfullytransparentregardinginput,andwehavenotbeen
abletoevaluatetheinterventionusingalternativeassumptions.Alsotheoutputisnot
reportedinadequatedetail.Thesubmitterhasreportedincrementalcosts,incremental
66 Cost-effectiveness analysis
effectsandICERs,whileabsolutelevelsofdirectcostsandutilitieswerenotreported,
whichisgoodpracticeincludingforvalidationpurposes.
Thesubmitterdidnotperformasensitivityanalysis,whichleadstolackofinformation
abouttheuncertaintyassociatedwiththedataincorporatedinthemodel.Thelackof
onewaysensitivityanalysesalsomakesitdifficulttoconsidertheinternalvalidityof
themodel.
The overall comments in the sections above regarding the low quality of the effec-
tiveness input data make the submiters model´s results very uncertain. The lack of
sensitivity analysis and the fact that we were not able to perform our own sensitivity
analysis due to de model constraints make the submitted results of cost effectiveness
(ICER) even less reliable.
Budgetimpactanalysis
Thesubmittercalculatedabudgetimpact,fromaNorwegianhealthcareperspective,
forintroducingFreeStyleLibreasasecond‐linetreatmentforindividualswithdiabetes
type1ortype2comparedtoSMBG.Thebudgetimpactwasestimatedasthenetcost
differencebetweenascenarioinwhichFreeStyleLibreisadoptedforafullcohortof
eligibleindividualsrelativetoascenarioinwhichthedeviceisnotadopted.Thebudget
impactwasestimatedastheyearlycostfiveyearsafteradoptionofthetechnology.
Thesubmittercreatedtwobudgetimpactscenarioanalyses,onescenarioanalysis
limitedtotype1DMonlyandanotherscenarioanalysiswhichincludedbothtype1DM
andtype2DM.Thesubmitterdidnotcreateabudgetimpactfortype2DMonly.
The submitted budget impact analysis
AccordingtodatafromtheNorwegianDiabetesAssociation2015(54),therewere
28,000individualswithtype1DMand8,220individualswithtype2DM,usinginsulin,
inNorway(54).Basedonthesenumbers,thesubmitterassumeda0.9%increasein
individualshavingdiabetestype1foreachyearanda3%increaseinindividuals
havingdiabetestype2eachyear.Table17belowshowsthepredictionofindividuals
withyear1asbeing2017asassumedbythesubmitter.Itstatestheannualnumberof
individualsusingFreeStyleLibreifthisnewtechnologywasadopted.Ifthenew
technologywasnotadopted,alloftheseindividualswouldreceiveSMBG(seeTable
17).Thesubmitterassumedthatanindividualwouldchangetheirsensorevery14th
day(41).
67 Cost-effectiveness analysis
Table17.Annualandincreasednumberofindividualswithadoptedtechnologyaccord‐
ingtosubmitter’smodel
Number of individuals if the new technology is adopted
Year 1 Year 2 Year 3 Year 4 Year 5
Annual number of individuals with type 1 DM 28,506 28,763 29,022 29,283 29,546
Increased number of individuals (type 1 DM) 0.9% 0.9% 0.9% 0.9%
Annual number of individuals with type 2 DM 8,721 8,982 9,252
Increased number of individuals (type 2 DM) 3% 3% 3% 3%
Annual number of individuals (type1 + type 2 DM) 37,227 37,745 38,273
Increased number of individuals (type 1 + type 2 DM) 1.4% 1.4% 1.4% 1.4%
Thecostoftype1and2DMwasestimatedbythesubmittertobeNOK239,515,116in
2015(55).ThesubmitterassumedthatthecostrelatedtoSMBGisequallydistributed
betweenindividualswithtype1and2DM.Further,thesubmitterassumedthatthe
costofFreeStyleLibrealsoincludesanadditionalcostof9%ofthetotalcostsrelatedto
SMGB.Thesubmitterassumedthatanincreasednumberofindividualswithbothtype
1and2DMwouldswitchfromSMBGtoFreeStyleLibreduringthefivefirstyears.
Thesubmittedbudgetimpactanalysisalsoincludesadditionalresourceuserelatedto
emergencyreception,ambulanceandhospitalstays.Numberofannualeventsperper‐
sonweretakenfromthetwotrials,IMPACT(25)andREPLACE(26).Theserates
showedmoreevents,whichledtomoreresourceuse,byusingSMGBonlyinsteadof
FreeStyleLibre.
ThesubmitterestimatedthatthetotaladdedcostswouldbeaboutNOK748millionfor
thefirstfiveyearsafteradoptionofFreeStyleLibreforindividualswithtype1DMin
Norway.Thesubmittercalculated,however,apotentialcostsavingofNOK75million
forthefirstfiveyearsafteradoptionofFreeStyleLibreforindividualswithtype1and
type2DMinNorway.Thesubmitter’sconcernsforconfidentialitydidnotallow
presentationoftheseresultsdisaggregatedly.
NIPH comments on the budget impact analysis
Thesubmitter’sbudgetimpactanalysiswasbasedonanExcel‐basedmodel,whichis
differentfromtheIMS‐CDMmodelusedforcost‐effectivenessanalysis.Thebudget
impactmodelincludesstudydataonresourceutilisationthatwherenotusedinthe
model.Basedonexpertopinions,theacuteeventswithseverehypoglycaemiaand
68 Cost-effectiveness analysis
ketacidosisappliesbothtoindividualswithtype1andtype2DMwhouseinsulin.
FreeStyleLibremayleadtomorestableglucoselevelsamongtheseindividualsbecause
ofmorefrequentbloodglucosemeasurement.Therefore,theseindividualsmay
decreasethenumberofemergencyreceptions,ambulanceandhospitalstays,andthen
reducetheresourceuserelatedtothesecomplications.
However,accordingtoourclinicalexperts,theadditionalcostrelatedtoSMBGwhile
usingFreeStyleLibremayhavebeenunderestimatedinthesubmittedcalculations.We
thereforerecalculatedthebudgetimpactanalysisbasedontheassumptionthat20%of
thetotalcostsrelatedtoSMBGismorereasonablethanthe9%appliedbythesubmit‐
ter.Webasedtherecalculationonthesubmitter’smodel,whichisExcelbasedandal‐
lowedustomodifyinputparameters.Further,ourexpertsassumedthat60%ofthein‐
dividualswithtype1DMandabout30%oftheindividualswithtype2DMinsulinus‐
erswouldstartusingtheFreeStyleLibreinsteadofSMBGonly.Ourclinicalexpertsas‐
sumedthatanindividualwouldlikelyhavetouse29sensorsperyear,asomewhat
highernumberthanfoundonthesubmitter’swebsite(41).Inouralternativeanalysis
wealsousedthepublicprice,NOK599(excl.VAT),foundonthesubmitter’sFreeStyle
LibrewebsiteJune,2017(41),sincethepricestatedinthesubmissionwasclaimedby
thesubmittertobeconfidential.
Basedonthesubmitters’modelandassumptionswecreatedthreebudgetimpact
scenarioanalyses,onescenarioanalysisrelatedtotype1DMonly,asecondscenario
analysisrelatedtotype2DMonly,andathirdscenarioanalysiswhichincludedboth
type1DMandtype2DMinsulinusers.Thesubmitterdidnotcreateabudgetimpact
fortype2DMonly.
Table18showsthebudgetimpactforindividualswithdiabetestype1.Thebudgetim‐
pactincludsthetwoscenarios:(1)costrelatedtoadoptionoftheFreeStyleLibreand
(2)costwithoutadoptionoftheFreeStyleLibre.Thecalculationsshowthedifference
betweenthetwoscenariosineachofthefiveyearsoftheanalysis.Thecomparisonsbe‐
tweenthetwoscenariosshowsanincreaseintotaladdedcostsforeachyear.Weesti‐
matedthatthetotaladdedcostswouldbeaboutNOK913,000,000forthefirstfive
yearsafteradoptionofFreeStyleLibreforindividualswithdiabetestype1inNorway,
oraboutNOK186millionperyearfiveyearsafterimplementation.
69 Cost-effectiveness analysis
Table18.NIPH’sbudgetimpactestimatesfortype1DM
Annual budget Impact
Year 1 Year 2 Year 3 Year 4 Year 5
+ Cost if the
New technology
is adopted
(NOK)
358,457,802 361,683,922 364,939,077 368,223,529 371,537,540
- Cost without
adoption of the
new technology,
i.e. current
situation (NOK)
179,074,804 180,686,477 182,312,655 183,953,469 185,609,051
Total added cost (NOK)
179,382,998 180,997,445 182,626,422 184,270,059 185,928,490
* Based on number of individuals estimated in Table 17; NOK: Norwegian Kroner
Table19showsthebudgetimpactforindividualswithtype2DMinsulinusers.Thecal‐
culationsshowthedifferencebetweenthetwoscenariosineachofthefiveyearsofthe
analysis.ThecomparisonsshowadecreaseintotalcostsofaboutNOK433,000,000for
thefirstfiveyearsafteradoptionofFreeStyleLibreforindividualswithtype2DMinsu‐
linusersinNorway,oradecreaseofNOK91,7millionperyearfiveyearsafterimple‐
mentation.
Table19.NIPH’sbudgetimpactestimatesfortype2DM
Annual budget Impact Year 1 Year 2 Year 3 Year 4 Year 5
+ Cost if the New
technology is adopted
(NOK)
29,801,725 30,695,777 31,616,650 32,565,150 33,542,104
- Cost without adoption of
the new technology, i.e.
current situation (NOK)
111,308,554 114,647,811 118,087,245 121,629,862 125,278,758
Total cost (NOK) (81,506,829) (83,952,034) (86,470,595) (89,064,713) (91,736,654)
* Based on number of patients estimated in Table 17; NOK: Norwegian Kroner
Inourthirdbudgetimpactscenarioanalysis,weassessedthetotaladdedcostof
introducingFreeStyleLibretobothindividualswithtype1DMandtype2DM.Based
onourcalculations,thetotaladdedcostwouldbethedifferencebetweenthetotal
addedcostsforindividualswithtype1DM(NOK913,000,000)andthetotalcost
savingsforindividualswithtype2DMinsulineusers(NOK433,000,000).Thetotal
70 Cost-effectiveness analysis
addedcostforintroducingFreeStyleLibretobothpopulationswouldbeaboutNOK
480million,oraboutNOK94millionperyearfiveyearsafterintroduction.
Asmentioned,thesubmitterappliedtwodifferentmodelsforthecost‐effectiveness
andbudgetimpactanalyses.Thesetwotoolsproducedsomeresultsthatare
counterintuitive,seenincombination.Mostimportantly,fortype2DM,thecost‐
effectivenessanalysisandtheIMS‐CDMproducedpositiveincrementalcostsfor
FreeStyleLibrecomparedtoSMBG(Table16),whilethebudgetimpactanalysisgave
theresultthatintroducingFreeStylelibreiscost‐savingforthesepatients.Inversely,
theIMS‐CDMproducednegativeincrementalcostsfortype1DM(Table15),whilethe
budgetimpactanalysisshowedsubstantialcostincreasesrelatedtointroductionof
FreeStyleLibre.Theseresultsinsumappearnon‐consistent.Whetherthisisduetothe
utilisationofdifferentdataindemodelandinthebudgetimpactcalculationsis
uncertain.
71 Discussion
Discussion
Wehaveperformedanindependentclinicalsystematicreviewandassessedthesub‐
mittedcost‐effectivenessinformationofFreeStyleLibreforindividualswithtype1and
2DM.WeconductedanindependentreviewoftheclinicalevidenceusingaPICO
framework;thePICOcomponentswereselectedincollaborationwiththeNorwegian
DiabetesAssociationandclinicalexperts.Ourmainoutcomesareclinicaloutcomes,
andwedidnotevaluatethesensitivityorspecificityofthedevice–i.e.,measuringhow
wellthedevicedoeswhatitissupposedtodo.
Clinicaleffectivenessandsafety:summaryofmainresults
Thesubmitter’sliteraturesearchidentifiedtworecords(oneRCTandoneposter).
ThesetworecordsreportedclinicaltrialsofFreeStyleLibreforindividualswithtype1
and2DMrespectively.Thesubmitterindicatedotherpublicationsprovidedusefulin‐
formationfortheaccuracyofthedevice(i.e.singlearmstudies).Thesubmitteralsoin‐
cludedanauthorityevaluationconductedinFrance(July2016)inthedocumentation
package(56).
WeconsideredthatonlytheRCTswererelevanttorespondtotheclinicaleffectiveness
andsafetyresearchquestion(25;26);theresultsofthesetrialswereusedinthemain
analysesinthisassessment.Weregardedthesubmitter’sinformationtobefairandthe
submitter’sinterpretationtobemostlyappropriate.Afterreviewingthe(smallamount
of)evidenceavailable,weconsideredthevalueplacedonthebenefitsofFreeStyleLi‐
brebypeoplewiththecondition,thosewhorepresentthem,andclinicalexperts.
Themeta‐analysisofFreeStyleLibreversusSMBGprovidedlowtoverylowqualityevi‐
denceonoutcomespostintervention.Therewerenodifferencesinseriousadverse
eventswithFreeStyleLibrecomparedwithSMBG.Therewerecleardifferencesinthe
effectsofFreeStyleLibreversusSMBGforindividualswithType1vstype2DMfor
someoutcomes.
72 Discussion
Traditionalmethodsofmeta‐analysissuggeststudiesarecombinedinoneanalysisif
theycomparesimilarpopulationsandsimilarinterventionsatsimilarfollowuptime
points,usingsimilaroutcomes.However,theanswertothefeasibilitytoconducta
meta‐analysisornotmayalsodependonthequestionbeingasked.
Themeta‐analysiswascarriedoutdespiteclinicaldifferencesinType1and2DM,as
webelievedthesynthesismatchedourresearchquestion.Ouranalysisaimedtoprove
thesubmitter’sclaim(“thetechnologyisindicatedformeasuringinterstitialfluidglu‐
coselevelsinpeople(age4andolder)withdiabetesmellitus”).AsGøtzche(57)points
outthereisadebatebetween“lumpig”and“splitting”resultsofclinicaltrialswhile
consideringmeta‐analyses.Clinicalexpertsandstakeholdersbelievedthecharacteris‐
ticsoftheindividualsdifferinawaytheydidnotrecommendtopoolthedatafromthe
studiesfound;inthiswaytheyadvocatedfor“splitting”.Afterfurtherconsultationwith
theNIPHteamstatisticianandexperts,wedecidedtocombinethedataaswebelieve
thisstatisticalapproachmatchestheresearchquestionbeingasked.Therationalwe
followedwhenwedecidedto“lump”theincludedRCTswasthatFreeStyleLibreshould
reflecttheobjectiveofthereport.Thisapproach,ofcourse,didnotpreventusfrom
lookingat,exploreandpresenttheresultsoranyresonsforheterogeneityinsub‐
groupswhennecessary.Nevertheless,thereadershouldbeawarethatthesynthesisis
basedontwosmallsubmitterinitiatedRCTs;thisinitselfbringsdifficultiesestimating
between‐studiesvariance,whichhasimplicationsformanyaspectsoftheanalysis.
Overall completeness and applicability of evidence
GiventhesmallnumberofRCTsincludedinthisassessment,webelieveitislikelythat
newtrialsmayaltertheestimatedeffectsofFreeStyleLibrefordiabetes.
RCTsincludedonlyadults,individualswithwellcontrolledtype1andpoorlycon‐
trolledtype2DM.Thus,ourconclusionsarelimitedtoadultswiththesecharacteristics
andfurthergeneralizationsarenotrecommended.
Weconsideredtheinformationregardingadverseevents(e.g.,safetyandharm)impre‐
ciseandofuncertainvalidity,primarilybecauseoftheincomparabletimesutilized,but
alsothesmallnumberofparticipantsinthesetrials.
Thenumberofstudiesisstillnotsufficientlylargetodrawfirmconclusionsontheclin‐
icaleffectivenessofFreeStyleLibre,noristheliteraturesufficientlyexplicitorlarge
73 Discussion
enoughtoanswerclinicalquestionsabouttheidealroutineorcombinationofFreeStyle
Librefor(all)individualswithDM.
WeacknowledgetheevidenceforFreeStyleLibreisincreasingandevolvingrapidily.
Weidentifiedanumberofongoingandcompleted(published/unpublished)studies.
Amongtheongoingstudiesthereare3RCTs:oneincludingyouth12to17years,an‐
otheradultswithtype2DM,andoneadultswithseverehypoglycaemia.Theresults
fromtheseongoingRCTswillaidourunderstandingoftheclinicaleffectivenessof
FreeStyleLibre,andwillprovidebetterinformationfordecisionmakersandthose
planningservicesforindividualswithdiabetes.
Quality of the evidence
Riskofbiasassessmentshighlightedconcernsregardinginsufficientinformationonal‐
locationconcealment,blindingofparticipantsandcareproviders,anddetectionbias
relatedtoselfreportedinstrumentsandoneobjectiveoutcome.
Otherissuestoconsiderwheninterpretingtheseresultsare:a)thestudiesassesseda
largenumberofoutcomes(morethanpresentedinthisreport),increasingtheproba‐
bilityoffindingstatisticallysignificantdifferencesforoutcomesbychance;andb)im‐
portantclinicalheterogeneitywaspresentamongthestudieswhichremainsamajor
challenge.
Theincludedstudiesweresponsoredbythesubmitter.Studieswereregisteredinthe
trialregistryforrandomizedcontroltrials,whichensuredpublicaccesstothefull
studyprotocol.Thesubmitters’roleandresponsibilitieswerefullydisclosedinthe
publishedstudies.However,financialrelationshipsamongindustry,scientificinvesti‐
gatorsandacademicinstitutionshavebeenseentoinfluenceresearchinimportant
ways(58).Industrysponsoredresearchisatopicofconcernthatiscontinuouslybeing
investigated.Theconcernrelatestoindustrysponsoredresearchbeingmorelikelyto
favourtheproductdevelopedbythecompanythanresearchfundedbyothersources
(59;60).Itwillbeimportantforfurtherassessmenttohaveindependentstudiesto
comparetheseresultswith.
Other European assessments
FreeStyleLibreassessmenthavebeenpublishedinotherEuropeancountries(seeTa‐
ble20).Oneoftheseassessmentswasincludedinthesubmitteddocumentationpack‐
74 Discussion
age(*)andtheotherswereretrievedbyoursearch.Overall,wefoundtheirresultssim‐
ilartoourassessment.Afterconsideringsimilaritiesinthemethodsandreporting,we
chosetocommentontheresultsofHauteAutoritédeSanté(HAS,France),andSESCS
(Spain).WethenbrieflycommentonTLVreportastheSwedishHTAassesementhas
notyetbeenpublished(e‐mailcorrespondence).
Table20.EuropeanHTAFreeStyleLibreassessments
Agency Title (date)
AQuAS – Agència de Qualitat i Avaluació Sani-
tàries de Catalunya, Spain (61).
Dispositiu Flash FreeStyle Libre® per al
monitoratge de la glucèmia (October 2016)
SESCS – Informes de Evaluacion de Technolo-
gias Sanitarias (62), Spain (translated document)
Efectividad, seguridad y coste-efectividad del
sistema flash de monitorizacion de glucosa en
liquido intersticial (FreeStyle Libre) para diabetes
mellitus tipo 1 y 2 (June 2016)
*HAS – Haute Autorité de Santé, France (56) Systeme FreeStyle Libre, Systeme flash d'auto-
surveillance du glucose (July 2016)
HTA-centrum Göteborg – Sweden (translated
document)
Kontinuerlig glukosmätning med FreeStyle Libre:
effekt på HbA1c hos typ 1 diabetiker (2015)
Agenas, Agenzia nazionale per i servizi sanitari
regionali, Italy Under appraisal
InadditiontotheaboveFreeStyleLibreassessments
ZIN–ZorginstituutNederland,TheNetherlands(63)Flash
glucosemonitoringsysteem(FreeStyleLibre)(February2016).Thisisan
individual’sclaimforfinancialcompensation.Theapplicantdidnotreceiveit
duetoFreeStylenotmeetingZIN’sscientificcriteria(translateddocument).
SwedishPaediatricSociety:Thedocumentdoesnotmention“FreeStyleLibre”,
butreferstoCGM.Itincludesasetofrecommendationsofuseofcontinuous
glucosemonitoringinchildrenandyouthwithtype1DM.(64).
TheDentalandPharmaceuticalBenefitsAgency(TLV)–Sweden(65)
TheFrenchnationalcommitteefortheevaluationofmedicaldevicesandhealthtech‐
nologiesreviewedFreeStyleLibreclinicaleffectivenessandaccuracyJuly2016.Intotal,
thereviewincludedtwonon‐RCTstudies(66‐68)fortheaccuracyandprecisionofthe
technology,andtwoRCTsnotpublishedattimeofevaluation(IMPACTandREPLACE
studies).HASrecommendedtheinclusionofthetechnologyasfollows:“Measurement
ofinterstitialglucoselevelsforthetreatmentofpatientswithtype1and2diabetes
75 Discussion
(adultsandchildrenaged4years)undergoingintensifiedinsulintherapy(usingexter‐
nalinsulinpumpor>3injectionsperday)andperformingtheSMBGseveraltimesa
day(>3perday).FreeStyleLibreisespeciallydesignedforindividualswhohavere‐
ceivedtherapeuticeducationandspecifictrainingontheuseoftheflashinterstitialglu‐
cosemonitoringsystem.”ThereportindicatesthattheinitialprescriptionofFreeStyle
Libremustbeensuredbyadiabetologistorapaediatricdiabetologist;theyfurtherin‐
dicatethereneedstobeaprovisionofatrialperiodofaminimumofonemonthforeli‐
gibleindividualsandthosewhocontinueusingFreeStyleLibreshouldundergoa3
monthevaluationtoassesswhetherornottocontinueusingthesystem.Inaddition,
priortoprescription,individualsshouldreceivespecificeducationtoprovidethem
withtheskillsandknowledgetoapplythesensorandtointerpreterandusetheinfor‐
mationprovidedbythesystem.Renewalisensuredbyanydoctor.Thecommission
emphasisedthatindividualcomfortandimprovedqualityoflifeduetolowercapillary
bloodglucosebyfingerpricktestimprovedwithFreeStyleLibre.HAS’assessmentpro‐
videsaguidanceforthereimbursementdecision.Nocost‐effectivenessassessment
analysiswascarriedoutforthisproduct(e‐mailcorrespondence)asitistheMinistry
ofHealththattakesthedecision.InMay2017theFrenchHealthMinistrygrantedna‐
tionalreimbursementofthetechnologyforinsulindependentindividualswithType1
and2DM.
TheassessmentbytheCanaryGovernment,SpainconductedinJune2016focusedon
theeffectiveness,safetyandcost‐effectivenessofFreeStyleLibreforindividualswith
type1and2DM.TheassessmentincludedoneRCTforindividualswithType1DM
(IMPACTstudy),andalsomentionstheREPLACEstudy,whichwasnotpublishedatthe
timethereportwasconducted.Researcherswerenotabletoidentifypublishedeco‐
nomicevaluations;thusAbbottprovidedone.Theagencyconcludedthereislimited
scientificevidenceregardingthesafetyandeffectivenessofFreeStyleLibre;similarlyto
ourassessment,theystatedtheavailablestudiesareindustryfundedandtheoverall
GRADEqualityoftheevidencewaslow.However,timespentinhypoglycemiawas
gradedamoderate.Alsoinagreementwithourreport,theassessmentmentionedthe
lackofseriousadverseeventsandrecommendstheuseofFreeStyleLibreshouldbe
donewiththeindividual,providingtrainingandeducationregardingadverseevents,
benefitsandrisks.Similarlytoourexperience,theeconomicassessmentpresentuncer‐
taintiesandlimitationsassociatedwiththeuseofcertainparametersandpointsoutto
thelackofasensitivityanalysis.Theagencystatestheresultsshouldbeinterpreted
76 Discussion
withcaution.Insummary,theagencymakesaconditionalrecommendationforFree‐
StyleLibreuseforindividualswithtype1DMwithcontrolledHbA1clevels(<7.5%),
andwithagoodprioradhesionintheuseofself‐monitoringglucoseinblood.
TheDentalandPharmaceuticalBenefitsAgency(TLV)inSweden,hasrecentlypub‐
lishedareportentitled“Nationalassessmentofmedicaldevicesforincreaseequityin
healthcare”(65).Thisreportdescribesaprocesswhichisabouttobeestablishedfor
assessingmedicaldevicesbeforedecidingonintroducingintotheSwedishhealthcare.
FreeStyleLibreisusedasexampleonhowamedicaldevicehasbeenintroducedbefore
anyassessmentonclinicaleffectivenessandcost‐effectivenesshasbeencarriedout.
Accordingtothereport,thishascausedsignificantbudgetimpactandledtounequal
accesstoFreeStyleLibretoindividuals,whichvarysignificantlybetweentheSwedish
regionstoday.PrescriptionsofFreeStyleLibrehaskeptonincreasingsincespring2016
inallregionsofSwedendespitethehighcosts,anditisestimatedthatuseofcontinu‐
ousglucosemonitors(suchasFreeStyleLibre)willleadtoatotalcostof600.000.000
SEKperyear.Therefore,inMay2016,TLVwasaskedtoassessthecost‐effectivenessof
continunousglucosemonitoring.Inthemeantime,theNewTechnologiesCouncil(NT‐
Rådet)haschosentoputonholdtherecommendationonusingFreeStyleLibreforin‐
dividualswithtype2DMuntiltheassessmentiscompletedinMay2017.TheNT‐Coun‐
cilisexpectedtogivetheirrecommendationthe7thofJune(69).ThefullSwedishHTA
reporthasnotbeenpublishedattimeofoursubmission(e‐mailcorrespondence).
Implications for clinical practice
WehaveusedtheEPICOTapproachtodescribingimplicationsofthepresentedevi‐
dence(70)
Evidence
Diabetesisachronicdiseasewithadurationofdecadesthatrequireslong‐termtreat‐
ment.Currentevidence(derivedfromtwoRCTs)isinadequatetoevaluatethebenefit
andharmsofFreeStyleLibreincomparisonwithstandardtreatment(SMBG).
Nevertheless,theevidenceseemstosuggestthatFreeStyleLibremayimprovetreat‐
mentsatisfaction,andslightlyimprovetimeinrange,glucose<3.9mmol/Lwithin24
hoursandnumberofevents,nocturnaleventsandtimespentwithglucose
<13.0mmol/LwhencomparedtoSMBG,howeverthequalityofevidenceislow.Indi‐
vidualsintheinterventionhadsimilarnumberofseriousadverseeventsasthoseinthe
77 Discussion
comparisongroup.Therewerenootherstatisticaldifferencesinotheroutcomesin‐
cludingqualityoflifeorchangeinHbA1c.Noneofthestudies,althoughconductedin
Europe,includedindividualsfromNorway.
Manystudiesconductedtodate(someprovidedbythesubmitter,retrievedbyour
search,andrecommendedbytheNorwegianDiabetesAssociation)wereexcludedbe‐
causetheydidnotmeetourPICOcriteria(eg,lackofacomparatorgroup).However,
withtheaimofpresentingacompletelistingofallFreeStyleLibreevidence,weprovide
informationonsinglearmcompletedstudies(publishedandunpublished)inAppendix
3andtrialregistryrecords(ongoingstudiesrecruitingornotyetrecruiting)inAppen‐
dix4.WesummarizedFreeStyleLibrecompletedandongoingstudiesintable21.
Table21.SummaryofcompletedandongoingstudiesforFreeStyleLibre
Population Completed(C)orOngoing(O)
RCTs Single arm
Type 1 DM – Children
Type 1 DM - Adolescent
--
1 (O)
3 (C) – 1 (O)
--
Gestational diabetes -- 1 (C)
Type 1 - adults 1 (C) 4 (C) – 1 (O)
Type 2 - adults 1 (C) – 1 (O) 1 (O)
Type 2-≥ 75 yrs of age -- 1 (O)
Type 1 & 2 - adults -- 4 (C) – 2 (O)
Adults - insulin dependent -- 1 (C)
Adults with severe hypoglycaemia 1 (O) --
Children, adults, seniors -- 1 (O)
Somesinglearmstudiespresentevidenceforpopulationgroupsnotaddressedbythe
includedRCTs.Forexample,threesinglearmtudies(66;71;72)includedchildren0to
18yearsold,andonestudyfocusedonType1and2pregnantwomen(73).Inaddition,
fourstudieshavefocusedexclusivelyonadultswithtype1DM(74‐77)andanother
fourincludedadultswithbothType1and2DM(68;78‐80).IshShalom(79)focusedon
individualswithdiabetesdifficulttocontrol.Oneunpublishedtrialregistryrecord(81)
doesnotspecifytypeofdiabetes(insulindependentadults).
78 Discussion
Simlarly,amongthetrialregistryrecords(orongoingstudiesrecruitingornotyetre‐
cruiting)wefoundevidencethatwillserveupdatingthisassessment.Therearethree
RCTs:onefortype1DMadolescents(82),oneinvolvingindividualswithtype2DM
(83),andthelastoneforindividualswithseverehypolgycaemia(84).Wealsofound7
singlearmstudies:oneinvolvingchildren(85),oneinvolvingindividuals4yrs+(chil‐
dren,adultsandseniors)(86),oneinvolvingadultswithType1DM(87)twoinvolving
adutlswithtype2DM(88;89),andtwoinvolvingadutlswithbothtype1and2DM
(90;91).
Population
Theempiricalevidenceincludedinthisreportislimitedtotwostudieswithlessthan
500whitemiddle‐agedadultparticipantslivinginEurope.Ourclinicalexpertsstressed
theinterpretationoftheresultsmustbelinkedtothecharacteristicsofthestudies’par‐
ticipants(e.g.,IMPACTstudyindividuals’HbA1clevels,REPLACEstudyindividuals’age
ofdiagnosisof40years).Therefore,thegeneralizabilityoftheresultstotheNorwegian
population/contextissubjecttodiscussion.
Althoughsinglearmstudieshaveinvolvedchildrenandpregnantwomentheseim‐
portantgroupswerenotincludedintheincludedRCTs.Also,thereisalackofevidence
(dervicedfromRCTs)forindividualswhomayrequirethesupportofacaregiverorex‐
periencedteamtousethetechnology.
Itwillbeimportanttoobtaininformationonlongtermuseofthetechnologyinchil‐
dren.TheinformationwillbeimportantasNorwayhasahighnumberofnewcasesof
type1diabetesinchildrenperyear(4).Accordingtooneoftheclinicalexpertscon‐
sulted,childrenandadolescentswilllikelybenefithighlyfromusingFreeStyleLibre.
Futurestudiesfocusingonchildren,pre‐teenagersandteenagersareimportant,asthis
cohortbehavesinadifferentwaythanadultsandFreeStyleLibreisafterall,abehav‐
iouralmanagementintervention.
Intervention
Individualswithdiabetes(and/ortheircaregivers)mustobtain,process,andunder‐
standbasichealthanddiabetesspecificdiseasemanagementinformationtomakeap‐
propriatedecisions.ThesubmittersuggestthepointofusingFreeStyleLibreistoin‐
creasethepatient’sglycaemicinsightanddiabetesmanagement.Weagreewiththe
submitterthattheinterventionmaycreateanopportunitytoencourageandempower
moresuitablediabetesmanagement.Butevidencepresentedinthestudiesislimitedto
79 Discussion
clinicaloutcomesandaselectedgroupofindividualswhohavelivedwiththedisease
formanyyearsandwerehabituatedto(self)managethedisease.Weagreetothe
uniquepotentialthetechnologyhastomakeanimpact,butcannotassumebasedon
thisassessmentorsubmitters’package,thetechnologyisbarrierfreeandadherence
andproperselfmanagementareguaranteed.Beforetechnologycanaidebehavioral
managementandadherence,ongoingdiabeteseducation,healthanddiabetesliteracy,
andsupportsystemsmayneedtobeestablishedsothatprovidersandindividualscan
usethetechnologysuccessfully.
Interactionsbetweenindividualsandhealthcareproviderspresentwithconsiderable
differencesindifferentcountries,whichwillaffectindividual’sbehaviourandtherefore
theeffectivenessofthetechnologyofinterest.Thus,theresultsfromtheincludedstud‐
iesmaynotberepresentativefortheNorwegiancontext.
ThesubmitterarguesthatFreeStyleLibremaychangediabetestreatmentas“itisthe
patientthatmakesthedaytodaydecisionsconcerningtreatment.”Treatmentdeci‐
sionswerenotassessedinthisreport.Forexample,theamountofinsulinperdayor
otherlifestylechangeslikedietandexercisewerenotconsidered,andthereforeweare
uncertainwhetherFreeStyleLibrecontributestochangetheseorotheraspectsofdia‐
betestreatment.
Notrainingwasprovidedtotheseparticipantsforinterpretationofglucosesensor
data.Thisisanimportantissueasonewouldexpectgreatereffectiftheindividualsin
theREPLACEstudyhadreceivedsuchtraining.Intheassessmentofcost,extraconsul‐
tationsareincluded,thusgivingthepossibilityforcorrecteducation.
Trainingandeducationaltoolsmaybeneededtoteachusersandtheircaregivershow
tomanageglucoselevels.Datainterpretationandcompetencemanagingglucoselevels
caninvolvealotoftimeanddedication.
Last,compatibilityregardingtechnologysoftwareandtechnologicalsystemsathome
forpeoplewithdiabetesandforhealthcareprofessionalsinNorwaycouldbeapointof
issue.
80 Discussion
Comparison
Thereiscurrentlyalackofevidenceforcomparisonsotherthancapillarybloodtesting.
Itisveryimportanttobeawarethatnoneoftheincludedstudiesusedcontinuousglu‐
cosemonitoring(withorwithoutcontinuoussubcutaneousinsulininfusion)asacom‐
parator.
Outcomes
QualityofLifeandTreatmentSatisfaction
QualityoflifeismentionedasoneofthemainfeaturesperceivedbyusersofFreeStyle
Libre.Ourassessmentshowedaneffectsizecrossingthelineofnoeffect,whichistell‐
ingusweareuncertainthatoneinterventionisbetterthantheotherforthisoutcome.
Theresultsweresurprisingforsomemembersoftheteam,butwerenotforsomeof
ourclinicalexperts.
Onepossibleexplanationfortheresultsisthattheparticipantswerealreadydedicated
tousenewtechnologies,andthelengthoftimetheindividualshavelivedwiththedis‐
easehasinfluencedtheirviewsonqualityoflife.Thisinadditiontoaperhaps‘short’
interventionnotallowingtoperceivefurtherimprovementsintheirqualityoflife.Also,
somestudieshavefoundthatbetterglycaemiccontrolisassociatedwithbetterquality
oflife,butotherstudieshavefailedtoshowthisrelationship(92;93).Oneofourclini‐
calexpertsadded,participantsrealizationofchangesinbloodglucosewithFreeStyle
Libre,evenwhentheydowhattheyaretoldbyhealthpracticioners,mayaffecttheir
qualityoflifeastheymayfeelnotbeingthemasterofthesituation.
Anotherpossibleexplanationisthetoolusedbyresearchers.TheDiabetesQualityof
Lifescalewasdevelopedforuseinthediabetescontrolandcomplicationstrial(DCCT)
tocomparetwotreatmentregimensforchroniccomplicationsinpatientswithType1
diabetes.Asoneofourstakeholderspointedout,thescaledoesnotseemtocoverim‐
portantpsychologicalaspectssuchasmotivationandcoping.Oneexplanation,isthat
thetoolisnotsensitiveenoughtomeasurechangesinthistypeoftrial.
Nevertheless,wefeltitwasimportattoaddthecommentofoneofthestakeholders
consulted(ayoungmanlivingwithdiabetesType1)ontheimpactofFreeStyleLibreon
hisqualityoflife.Hetoldus:
81 Discussion
Thelifeofadiabetes1patientischaracterizedbyahighdegreeoflimitationsand
FreestyleLibretakesawaysomeoftheselimitations(e.g.,soccertraining,swim‐
ming,skiingandatschool).Whenyouhaveachronicdiseasethatyouwillhave
for70‐80years,evensmallimprovementsinqualityoflifehaveamajorimpactin
thelongrun!!!
Treatmentsatisfaction,however,hadachangeof5points(onascale‐18to18points)
which,accordingtoourclinicalexperts,isclinicallysignificantimprovement.Typeof
insulinregimen,thefrequencyofbloodglucoseself‐monitoringandperceivedfre‐
quencyofhyperglycaemiacanhaveasignificantassociationwithtreatmentsatisfac‐
tion.Onecanarguethattreatmentsatisfactionispartofqualityoflife,thusconfirming
thatFreeStyleLibreimprovesthelivesofindividualswithdiabetes.
HbA1c(%)
ThesubmittersuggestedtheREPLACEandIMPACTtrialsweredesignedtoreflectrou‐
tineclinicalpracticeandthattheendpointsusedinthetrialswererelevant,inparticu‐
larHbA1c.Wedisagreewiththesubmitter’sstatement,andasoneofourclinicalex‐
pertssuggested,thepopulationinthetrialsmayhavebeenselectedpurposefully
Inotherwords,intheIMPACTtrialpopulationcharachteristics(i.e.,wellcontrolled
HbA1c)itwaslikelyresearcherswouldfindchangesinhypoglycemia(mainendpoint
ofthestudy).Similarly,wefoundednoteworthymaingoaloftheREPLACEstudy
(changesinHbA1c)andtheparticipants’characteristics(i.e.,notwellcontrolled
HbA1c).
TherewasevidenceofalackofaneffectforHbA1c,however,thiswasnotasurprising
result.IndividualsintheIMPACTstudywerewellcontrolled,withameanlevelof6.7%
(25).Oneofourclinicalexpertarguedthattheindividual’sdiseasewasextremelywell
regulatedinthisstudy,anditwouldevennotberecommendedforthemtoreducetheir
HbA1c<6.5%.Participantsinthisstudywereratherdedicatedandmeasuredblood
sugarregularlyandoften.Itislikelythatiftheywerenotthatwellregulated,itwould
bemorelikelytoseeareductioninHbA1c.ForindividualswithHbA1cof7.5%itwould
benecessarytoreducetheirHbA1c.Thisscenariohardlyreflectsclinicalpractice.
IndividualswithType2DMweredifferent,andlesspronetohypoglycaemia.Thegroup
studiedwasnotwellregulatedwithanHbA1cof8.8%(Haak),indicatingthattheyeven
fromthebeginningofthestudywerenotdedicatedindividuals.
82 Discussion
Inaddition,thestudybyHaaketal.includedolderparticipants,meanage59years.
Participantsyoungerthan65yearsofagehadalargerfallinHbA1clevelintheinter‐
ventiongroupcomparedwiththecontrolgroup.FurthermorethemeanBMIwasvery
high,33.1,indicatingthatseveraloftheparticipantshadametabolicsyndromeeven
thoughthosewithinsulinconsumptionover1.75units/kg/daywereexcluded.Those
withpoordiabetescontrolandsignsofmetabolicsyndromeneedcombinedinterven‐
tionstoachievegooddiabetescontrol.Thesubmitterindicated“Afocusonprevention
ofhypoglycaemiainthisgroupmayexplainwhyHbA1clevelswerenotsignificantlyre‐
duced”butwehavenotindicationofwhatthismeans.
Thishighlightsthedifficultiesinidentifyingandincludingindividualsforthetrialsrele‐
vanttothetechnology.Basedonthis,weareunsurehowFreeStylewouldperformon
individualswithpoorlycontrolledHbA1c.Thesubmitteracknowledgeslimitationsof
thetechnologyforsituationsofrapidlychangingglucoselevels,confirmationofhypo‐
glycaemiaorimpendinghypoglycaemia,orwhensymptomsdonotmatchFreeStyleLi‐
brereading.Asoneofourclinicalexpertsmentioned,thesearesituationswhenindi‐
vidualsareingreatestneedofknowingexactlywhattheirpresentglucoselevelis.
Theselimitationshavetobetakenintoconsideration.
Pain
Wehavenotbeenabletoreportonthisoutcomeasitwasnotdirectlymeasuredinthe
studies.Individuals’adherencetoinsulintherapyisanongoingchallengeinclinical
care.Painisobviouslyaveryrelevantoutcome,inchildreninparticular,andareason
forchoosingonetreatment(ortypeoftechnology)overanother.Injection/needlere‐
lateddiscomfortcontinuestobearheavilyintheindividual,thecaregiverandsome‐
timesthehealthcareprovider.
Glycaemiccontrol
Wedonotagreewiththesubmitters’claimthattheusersofFreeStyleLibredecreased
mostaspectsofhypoglycemia.Resultsofouranalysisshowonlyhalfoftheoutcomes
werestatisticallysignificant.
Thereweresomeclear,andexpected,differencesinglycaemiccontroloutcomesfor
type1and2DM.Asaclinicalexpertpointedout,intype1DMamorerapidriseandfall
inbloodglucoseisobservedsincetheseindividualsdonotproduceanyinsulinand
thusdependcompletelyoninsulinsuppliedsubcutaneously.Itisthereforenotso
strangethat,whenlookingattheforestplots,theeffectofFreeStyleLibreisgreatestin
83 Discussion
thetype1DMgroup.Whilelookingatthepooledresults,ouranalysisshowednostatis‐
ticallysignificantresultsfortimeandeventswithglucose<3.1mmol/L,and<3.1
mmol/Latnight.Thiswasalsothecaseforhyperglycaemiatime>10mmol/L(type
2DM).Aclinicalexpertmentionedareductioninhypoglycaemia(especiallyinthelow
group<3.1mmol/L)isofclinicallysignificance,andtheimportanceofthisasitmayre‐
ducemorbidityandmortalityduetolossofconsciousnessandsecondarycardiovascu‐
larincidents.Further,ourresultsshowedonlyareductionofnumberofeventswith
glucose<3.9mmol/Lwithin24hours.
Adverseevents
Wedisagreewiththesubmitter’sfavourableinterpretationofadverseeventsrelatedto
thedeviceandwithdrawalscomparedwithSMBG.Adverseeventsweremeasuredat
uncomparabletimes(i.e.,baselineand6monthsendpointfortheinterventiongroup,
andbaselineand4weeks‐14daysat3monthsand14daysat6month‐inthecontrol
group)constitutingamajorthreatintheinterpretationoftheresults.Assuch,thecur‐
rentresultsmaybemisleadingandneedtobereadwithcaution.Inaddition,thelow
numberofparticipantsandtheshortlengthoftheinterventiondonotjustify,inour
opinion,suchoptimisticconclusions.
Issuesrelatedtothetechnology,likescaringorirritabilityoftheskinintheareas
wheresensoriswornmayneedtobetakenintoconsiderationasalternativeareasto
sensorwearingasitwasneededinatleastoneoftheseRCTs(25).Thetrialperiodis
notlongenoughtounderstandtheassociationsbetweenconsequencesinsubcutane‐
ousskinandintersticialreading.
Somesensorsfailandconsistentsupplyofsensorsisnecessary;thiscancausetheuser
tohavetouseothermethodsformanagementofthedisease.Asoneofourclinicalex‐
pertsmentioned,individualswillstillneedsomeSMBGtocheckiftheFreeStyleshow
correctvalues.
Non‐devicerelated,device‐relatedandwithdrawalsfromthestudywerenotedinboth
trials.Inagreementwithourclinicalexperts,weconcludedthatFreeStylehasatolera‐
bleshorttermadverseeventprofile,enphasazingthateventhatinformationhastobe
takenwithcaution.
84 Discussion
Oneofourclinicalexpertsmentionedthatindividual’sfatigueinwearingtheFreeStyle
Libreisnottakenintoaccount.Neitheristhefactthatinitialstudiesinparticularlyded‐
icatedpatientstendtooverestimatetreatmenteffectscomparedtowhenthetreatment
isusedinageneraldiseasepopulation.
Limitationsoftheassessment/technology
Themainuncertaintiesregardingclinicaleffectivenessarethegenerallackofdata
(includingchildren,youth,andspecialpopulations)andlowqualityofavailable
data.
Duetoinheritdifferenceindiseasecharacteristicsitwasdifficulttoevaluateand
interpretthedatainthisassessment.
Oneoftheclinicalexpertssuggestedthereisaveryclearrelationshipbetween
timespentinhypoglycaemiaandriskofseverehypoglycaemia.Thisis
especiallyimportantinindividualswithimpairedawarenessofhypoglycaemia.
Theseindividualswill/shouldoftenchoseatechnologywithanalarmthatwill
indicatetheglucoselevelisapproachinghypoglycaemia.FreeStyleLibreis
unabletoprovidesuchalarm,andthisisthereforealimitationofthe
technology.
TheFreeStyleLibreglucosesensorissituatedundertheskinandtherefore
measurestheglucoselevelsinthefluidbetweenbloodcapillariesandthe
body’scells(interstitialfluid)ratherthancapillarybloodglucoselevels.
Becauseglucosemovesfromthecapillariestotissues,thereisalagbetween
bloodandinterstitialglucoselevelsofatleast6‐7minutes,butontopofthis
comesadelayinthesensor(94).Thelagincreaseswhenglusoeisincreasing
anddecreaseswhenglucoseisdecreasing,andtheeffectislargerwhenthe
speedofchangeinglucoselevelsarelarger.Toknowthedelayandtheabso‐
lutedifferenceonemustdotestwiththeactualglucosesensor.Further,
confirmatorycapillarybloodglucosetestsmaythereforeberequiredtoconfirm
thevaluedisplayedbythecontinuousglucosemonitorbeforemakingany
adjustmentstodiabetestherapyordrivingavehicle.
Cost-effectiveness
Thesubmitterprovidedacost‐effectivenessanalysisoveratimehorizonof40‐years.
Thesubmittercalculatedanincrementalcost‐effectivenessratioforFreeStyleLibre
comparedwithSMBG.Thehealtheconomicevaluationwasconductedusingaweb‐
basedmodel,beingtheIMSCDM.ThemodelwasbuiltuponMarkovmodellingof17
sub‐modelsofcomplicationsassociatedwithType1and2DM.
85 Discussion
Wehavehighlightedsomeweaknessesinthemodelthatwethinkshouldbeaddressed.
First,themodelreceivedfromthesubmitterlackstransparency,makingitdifficultto
gainafirmunderstandingofthefactorsthatdetermineacohort’sprogressthroughthe
model,assumptionsandparameterseffectoutcomesandtoassessthevalidityofthe
model.Thismadeusunabletoevaluatethemodelstructure,andtrackwhat’sinflu‐
encedtheresults(internalvalidation).Themodelhasbeenconsideredforitsexternal
validity,latestperformedbyMcEwanetal.2014(34).AccordingtoMcEwanetal.2014
(34)areviewmethodusedinlongtermcost‐effectivenessmodelsofdiabetesinterven‐
tionsalsoconfirmedtheprominentrolethatthediabetescontrolandcomplications
trialandtheUnitedKingdomProspectiveDiabetesStud(31;33)playinprovidingdata
onefficacyanddiseaseprogression,particularlyfordefiningtherelationshipbetween
HbA1candcomplications.
However,someissuesarenotsufficientlycommentedbythesubmitter.First,the
modelandvalidationfocusonthedisease,diabetes,butnotparticuarlythetechnology
andprocedurerepresentedbyFreeStyleLibre.Inputisthusnotvalidated.Secondly,
thesubmittedmodelshowsanimprovementinHbA1cforuseofFreeStyleLibrecom‐
paredtobloodglucosemonitoring.Thisisnotthecasewithourefficacyfindingsinour
singletechnologyassessment,whichshowsthattherewereapproximatelynodiffer‐
enceinHbA1cfortheinterventionversusthecomparator.Thesubmitterhasbasedthe
resultson6monthsdata,whichisaveryshortperiodoftime,asHbA1cissomething
thatinfluencesdiabetesthroughalifetimeperiod.Evidenceforalongertimeperiodis
notpresent,whichleadstolargeuncertaintyabouttheresults.Third,thesubmitterdid
notprovidereferencestoseveralsourcesoftheinputdataandqualityoflifeimprove‐
ments.Forexample,theutilityvaluerelatedtotype1DM,withnocomplicationsof0.9.
Inaddition,thesubmitterusedifferentvaluesintheefficacypartandinthemodel.Re‐
gardingthecosts,thepriceofFreeStyleLibrewerebasedonthepricefoundontheoffi‐
cialweb‐siderelatedtothedevice(41),whichisanotherpricethanusedbythesubmit‐
ter.Thepriceandtheothercostsdidnotincludevalueaddedtax(VAT).VATshould
havebeenincludedinthisanalysissinceitappliestheperspectiveofthehealthcare
provider.
Basedoninputgivenbytheclinicalexpertsandthoroughreview,webelievethatthe
healtheconomicmodelcapturedtheoutcomesthatareclinicallyrelevantforthede‐
finedpopulationandintervention.Accordingtoourclinicalexperts,thestudypopula‐
tioninIMPACTandREPLACEweresaidtobecomparabletotheNorwegianpopulation.
86 Discussion
Ontheotherhand,variousinputdatainthemodelistakenfromstudieswithdifferent
inclusioncriteria(forexampleaboutthestudypopulationanddiabetestype),which
leadstoagreatuncertaintyabouthowwellthedatafittotheuseofFreeStyleLibrein
Norway.Theclinicalinputdatarelatedtoindividualswithtype1DMwerebasedon
differentreferences.InonesimulationtheirclinicalinputswerebasedontheREPLACE
trial,whichincludespatientswithtype2DM,under65years.Inanothersimulation
theincludedclinicalinputdatawerebasedonSwedishfindings.Severaloftheclinical
inputdataincorporatedinthesubmittedmodel,didnotcorrespondtotheclinical
inputdatalistedupinthesubmitteddocument.Further,itisuncertainhowcompara‐
bletheinputdataaretoeachother,sincemostofthecomplicationsarefromseparate
studies.ThetimehorizonintheIMPACTandREPLACEtrialswasonly6months,andin
themodeltheoutcomesareextrapolatedover40yearsassumingcontinuousbenefits.
Thecalculatedincrementalcost‐effectivenessratio(ICER)basedonthesubmittedeco‐
nomicmodelovera40‐yeartimehorizonwasdominantforpatientswithtype1diabe‐
tes,meaningthatFreeStyleLibreaccordingtothemanufacturerisbothcostsavingand
moreeffective.Forpatientswithtype2diabetestheICERwascalculatedtobeNOK
235,673perQALY(wholestudypopulation)andNOK243,434perQALY(under65
years).Asthesubmittedmodelwasnottransparentandnotsufficientlyflexible,we
werenotabletorunamodelwithadjustedinputdata.
ThesubmitterestimatedthatFreeStyleLibrewouldbepotentiallycostsavingthefirst
fiveyearsafteritsadoptionforpatientswithdiabetestype1anddiabetestype2insu‐
lineusersinNorway.However,weadjustedsomeofthecalculationsrelatedtothe
budgetimpactanalyses,basedoninputdatafromourclinicalexperts.Ourresultsindi‐
catedthatthetotaladdedcostswouldbeaboutNOK913millionthefirstfiveyearsaf‐
teradoptionofFreeStyleLibreforonlypatientswithdiabetestype1,andaboutNOK
480milliontotaladdedcoststhefirstfiveyearsafteradoptionofFreeStyleLibrefor
patientswithdiabetestype1anddiabetestype2insulineusers.Theannualcostsfive
yearsafterintroductionwouldbeNOK186millionaddedcostandNOK91,7million
savedcostfortype1and2DMalone,respectively,andNOK94millionaddedcostfor
type1and2DMcombined.
Wefounditstrangethatthesubmitter’scost‐effectivenessresultsandbudgetimpact
resultsdeviatesfromeachother,i.e.adominantedICERandahightotaladdedcostin
thebudgetimpact.
87 Discussion
Nosensitivityanalyseswereprovidedbythesubmitter,whichisacrucialpartofan
economicevaluation,anditwasthereforenotpossibletoassesshowrobusttheresults
aretovariationinuncertainsingleparameters.Wecanthusnotseehowpricesand
otheraspectsmightinfluencetheresults.Onewaysensitiveityanalysesareinaddition
importantforconsideringtheinternalvalidityofthemodel.Thesubmitterhasonly
doneonescenarioanalysiswithasmallpricedifference.
KeyConclusions
Alloftheexistingdiabetestechnologieshaveadvantagesanddisadvantages,andperse
noneenableeveryonewithdiabetestoachievetargetHbA1clevels.FreeStyleLibre
maybeusedtoinformindividualsregardingtheirinterstitialglucoseleveltoallowde‐
cisionmakingregardingtreatmentandtogivethepersoninformationontheimpactof
changesinlifestyle,dietandphysicalactivity,whichcouldinfluencethelong‐termcon‐
troloftheirdisease.
Efficacy
ThereislimitedandlowqualityscientificevidenceontheclinicalefficacyofFreeStyle
Libre;thisbringslimitationstothegeneralizationoftheresults.Thetwotrialswehave
evaluatedindicatethatFreeStyleLibremayimprovetreatmentsatisfaction,timespent
inhypoglycaemiaandnumberofhypoglycaemicevents,butthestudiesweretoofew
andsmalltosupporttheclinicalefficacyofFreeStyleLibretothebroadercommunity
ofindividualswithdiabetes.Inaddition,thestudiesincludedrepresentonlyapartof
thepopulationofpeoplewithdiabetesinNorway,anditisthereforedifficulttomake
generalizationstotheNorwegiancontext.
FreeStyleLibremayslightlyimprovetreatmentsatisfaction,timeinrange,numberof
nocturnalevents<3.1mmol/Lwithin7h,andtimewithglucose>13.0mmol/Lincom‐
parisontoSMBG;FreeStyleLibreleadtolittleornodifferenceinqualityoflifeand
HbA1cincomparisontoSMBG.WeareuncertainwhetherFreeStyleLibreleadstoan
improvementintimeandeventswithglucose<3.9mmol/Lwithin24h,timewithglu‐
cose<3.1mmol/Latnightwithin7hours,andtimewithglucose>10mmol/L.Thisap‐
pliesonlytoindividuals18yearsorolder,andindividualsunder65yearswithtype2
DM,aswellasindividuals18yearsorolderwithtype1DM
OursearchfoundseveralFreeStyleLibreregistryrecordsofongoingstudies(seeAp‐
pendix4).Amongtheseongoingtrials,threehavearandomizedcontrolleddesignand
88 Discussion
canfurtherourknowledgeontheclinicaleffectivenessofFreeStyleLibre.Theresultsof
thisassessmentwilllikelychangewhennewevidencebecomesavailable.
Cost‐effectiveness
Theresultsofsubmitters’analysisshowedthatFreeStyleLibreisacost‐effectivealter‐
nativetoSMBG.However,wewerenotabletocontrolthesubmittedresultsandto
makeourowncost‐effectivnessresults(ICERs).Thereareseveralproblemswiththe
modelthatleadustoquestionthesubmittedresult.Themostchallengingissueisthat
themodelisnottransparent.Otherchallengingissuesaretheshorttermdatausedin
themodel,whichshouldcapturethedifferenceinefficacybetweentheFreeStyleLibre
usersandindividualswhoareusingSMBGonly.Also,thesubmittedmodelincluded
severalinputdatathatneithermatchedtheinputdatadescribedinthesubmitteddocu‐
mentationpackagenormatchedtheinputdatafoundinotherliterature.
Becausewedidnothaveaccesstothecompletemodelwewerenotabletoassessthe
validityofthesubmitters’estimates,norhowadjustmentofthemodelwithalternative
assumptionswouldaffecttheresultsprovidedbythesubmitters.Takinginaccountthe
commentsabovetheNIPHfindstheICERestimatesunreliable.TheNIPHistherefore
notabletoconsiderwhetherornotFreeStylelibreiscost‐effectivebasedonthesub‐
mitteddocumentation.
AccordingtothebudgetimpactanalysisperformedbyNIPHtheannualcostsfiveyears
afterintroductionwouldbeNOK186millionaddedcostandNOK91,7millionsaved
costfortype1and2DMalone,respectively,andNOK94millionaddedcostfortype1
and2DMcombined.
OverallFreeStyleLibredoesnotseemtoprovideahigherefficacyorfeweradverse
eventsorincreasedqualityoflifemeasuresforinsulintreatedpatiensthanotherSMBG
devices,thusitmakesdifficulttosupportthelowercostsassociatedwithFreeStyleLi‐
bre.
89
Appendices
References
1. DanaeiG,FinucaneMM,LuY,SinghGM,CowanMJ,PaciorekCJ,etal.National,regional,andglobaltrendsinfastingplasmaglucoseanddiabetesprevalencesince1980:systematicanalysisofhealthexaminationsurveysandepidemiologicalstudieswith370country‐yearsand2.7millionparticipants.Lancet2011;378(9785):31‐40.
2. ForouhiNG,WarehamNJ.Epidemiologyofdiabetes.Medicine2014;42(12):698‐702.
3. WorldHealthOrganization.Noncommunicablediseases.Diabetes.DataandstatisticsDenmark:WHO[citedMar31].Availablefrom:http://www.euro.who.int/en/health‐topics/noncommunicable‐diseases/diabetes/data‐and‐statistics
4. SteneL,StrømH,GulsethH.DiabetesinNorway‐PublicHealthReport2014.Oslo,Norway:NorwegianInstituteofPublicHealth2014.Availablefrom:https://www.fhi.no/en/op/public‐health‐report‐2014/health‐‐disease/diabetes‐in‐norway‐‐‐public‐health‐/
5. ThomasERA,BrackenridgeA,KiddJ,KariyawasamD,CarrollP,ColcloughK,etal.Diagnosisofmonogenicdiabetes:10‐Yearexperienceinalargemulti‐ethnicdiabetescenter.JournalofDiabetesInvestigation2016;7(3):332‐7.
6. NomiyamaTFauYT.[Secondarydiabetes].JapaneseJournalofClinicalMedicine2015;73(12):2008‐12.
7. HannaFW,DuffCJ,Shelley‐HitchenA,HodgsonE,FryerAA.Diagnosinggestationaldiabetesmellitus:implicationsofrecentchangesindiagnosticcriteriaandroleofglycatedhaemoglobin(HbA1c).ClinMed(Lond)2017;17(2):108‐13.
8. AmericanDiabetesAssociation.Whatisgestationaldiabetes?Arlington,VA[citedMay8].Availablefrom:http://www.diabetes.org/diabetes‐basics/gestational/what‐is‐gestational‐diabetes.html?referrer=https://www.google.no/
9. DiabetesAustralia.GestationalDiabetesCanberra,ACT:AustralianGoverment[citedMay8].Availablefrom:https://www.diabetesaustralia.com.au/gestational‐diabetes
10. McGavockJ,DartA,WicklowB.Lifestyletherapyforthetreatmentofyouthwithtype2diabetes.CurrDiabRep2015;15(1):568.
11. HamiltonH,KnudsenG,VainaCL,SmithM,PaulSP.Childrenandyoungpeoplewithdiabetes:recognitionandmanagement.BrJNurs2017;26(6):340‐7.
12. KiseSS,HopkinsA,BurkeS.ImprovingSchoolExperiencesforAdolescentsWithType1Diabetes.JSchHealth2017;87(5):363‐75.
13. StGeorgeSM,PulgaronER,FerrantiD,AgostoY,ToroMI,RamseurKC,etal.AQualitativeStudyofCognitive,Behavioral,andPsychosocialChallenges
90
AssociatedWithPediatricType2DiabetesinEthnicMinorityParentsandAdolescents.DiabetesEduc2017;43(2):180‐9.
14. SørensenM,ArnebergF,LineTM,BergTJ.CostofdiabetesinNorway2011.DiabetesResClinPract2016;122:124‐32.
15. GuptaN,BhadadaSK,ShahVN,MattooSK.PsychologicalAspectsRelatedtoDiabetesMellitus.JournalofDiabetesResearch2016;2016:7276403.
16. Young‐HymanD,deGrootM,Hill‐BriggsF,GonzalezJS,HoodK,PeyrotM.PsychosocialCareforPeopleWithDiabetes:APositionStatementoftheAmericanDiabetesAssociation.DiabetesCare2016;39(12):2126‐40.
17. GravelingAJ,FrierBM.Impairedawarenessofhypoglycaemia:areview.DiabetesMetab2010;36(Suppl3):S64‐74.
18. KennyC.Whenhypoglycemiaisnotobvious:Diagnosingandtreatingunder‐recognizedandundisclosedhypoglycemia.PrimCareDiabetes2014;8(1):3‐11.
19. AustadB,HetlevikI,MjølstadBP,HelvikA‐S.Applyingclinicalguidelinesingeneralpractice:aqualitativestudyofpotentialcomplications.BMCFamPract2016;17(1):92.
20. HigginsJPT,AltmanDG,SterneAC.Assessingriskofbiasinincludedstudies.In:GreenJHaS,editor.CochraneHandbookforSystematicReviewsofInterventions.Version5.1.0ed.www.handbook.cochrane.org:TheCochraneCollaboration;2011.
21. ReviewManager5.3[computerprogram],.Copenhagen:TheNordicCochraneCentre,TheCochraneCollaboration;2014.
22. DeeksJJ,HigginsJPT,AltmanDG.Analysingdataandundertakingmeta‐analysis.In:GreenJHaS,editor.CochraneHandbookforSystematicReviews.www.handbook.cochrane.org:TheCochraneCollaboration;2011.
23. GRADEproGDT.GRADEproGuidelineDevelopmentTool[Software].McMasterUniversity;2015(developedbyEvidencePrime,Inc.).
24. AbelsonJ,WagnerF,DeJeanD,BoesveldS,GauvinF‐P,BeanS,etal.Publicandpatientinvolvementinhealthtechnologyassessment:AframeworkforactionIntJTechnolAssessHealthCare2016;32(4):256‐64.
25. BolinderJ,AntunaR,Geelhoed‐DuijvestijnP,KrogerJ,WeitgasserR.Novelglucose‐sensingtechnologyandhypoglycaemiaintype1diabetes:amulticentre,non‐masked,randomisedcontrolledtrial.TheLancet2016;388(10057):2254‐63.
26. HaakT,HanaireH,AjjanR,HermannsN,RivelineJP,RaymanG.FlashGlucose‐SensingTechnologyasaReplacementforBloodGlucoseMonitoringfortheManagementofInsulin‐TreatedType2Diabetes:aMulticenter,Open‐LabelRandomizedControlledTrial.DiabetesTherapyResearch,TreatmentandEducationofDiabetesandRelatedDisorders2016;20:20.
27. BrahimiN,PotierL,MohammediK.CutaneousadverseeventsrelatedtoFreeStyleLibredevice.TheLancet2017;389(10077):1396.
28. DrummondMF,SculpherMJ,TorranceGW,O'BrienBJ,StoddartGL.Methodsfortheeconomicevaluationofhealthcareprogrammes.4thed.ed.Oxford:OxfordUniversityPress;2015.
29. BilirSP,LiH,WehlerEA,HellmundR,MunakataJ.Costeffectivenessanalysisofaflashglucosemonitoringsystemfortype1diabetes(T1DM)patientsreceivingintensiveinsulintreatmentinEuropeandAustralia.ValueHealth2016;19(7):A697‐A8.
30. LiH,BilirSP,WehlerEA,HellmundR,MunakataJ.Costeffectivenessanalysisofaflashglucosemonitoringsystemfortype2diabetes(T2DM)patientsreceivingintensiveinsulintreatmentinEurope.ValueHealth2016;19(7):A698.
31. DCCT.Theabsenceofaglycemicthresholdforthedevelopmentoflong‐termcomplications:theperspectiveoftheDiabetesControlandComplicationsTrial.Diabetes1996;45(10):1289‐98.
91
32. HoKK,AndersonKM,KannelWB,GrossmanW,LevyD.SurvivalaftertheonsetofcongestiveheartfailureinFraminghamHeartStudysubjects.Circulation1993;88(1):107‐15.
33. ClarkePM,GrayAM,BriggsA,FarmerAJ,FennP,StevensRJ,etal.Amodeltoestimatethelifetimehealthoutcomesofpatientswithtype2diabetes:theUnitedKingdomProspectiveDiabetesStudy(UKPDS)OutcomesModel(UKPDSno.68).Diabetologia2004;47(10):1747‐59.
34. McEwanP,FoosV,PalmerJL,LamotteM,LloydA,GrantD.ValidationoftheIMSCOREDiabetesModel.ValueHealth2014;17(6):714‐24.
35. FoosV,PalmerJ,GrantD,LloydA,LamotteM,McEwanP.TheIMSCOREDiabetesModelUserGuide.2.0ed.Switzerland:2012.AGuidetoPreparingandPerformingAnalysesUsingtheIMSCOREDiabetesModel.
36. PalmerAJRS,ValentineWJ,MinshallME,FoosV,LuratiFM,LammertM,SpinasGA,.TheCOREDiabetesModel:ProjectingLong‐termClinicalOutcomes,CostsandCosteffectivenessofInterventionsinDiabetesMellitus(Types1and2)toSupportClinicalandReimbursementDecisionmaking.CurrMedResOpin2004;Aug;20Suppl(Suppl1):S5‐26.
37. BuseJ,BergenstalR,GlassL,HeilmannC,LewisM,KwanA,etal.UseofTwice‐DailyExenatideinBasalInsulin–TreatedPatientsWithType2DiabetesARandomized,ControlledTrial.AnnInternMed2011;154(2):103‐12.
38. HellerSR,ChoudharyP,DaviesC,EmeryC,CampbellMJ,FreemanJ,etal.Riskofhypoglycaemiaintypes1and2diabetes:Effectsoftreatmentmodalitiesandtheirduration.Diabetologia2007;50(6):1140‐7.
39. Statistisksentralbyrå.Konsumprisindeksen,mars2017Norway[cited].Availablefrom:https://www.ssb.no/kpi
40. Statenslegemiddelverk.Maksimalpris:prisoversiktper01.05.2016(Excel)[cited01.03.2017].Availablefrom:http://www.legemiddelverket.no/Blaa_resept_og_pris/pris‐paa‐legemidler/maksimalpris/Sider/default.aspx
41. AbbottDiabetesCare.FreeStyleLibreFlashGlucoseMonitoringSystem:AbbottNorgeAS[updated01.01.2014;cited29.06.2017].Availablefrom:http://www.freestylelibre.no/
42. Apotek1.FreeStylePrecisionTeststrimmel50stkNorway[cited01.03.2017].Availablefrom:https://www.apotek1.no/produkter/freestyle‐precision‐teststrimmel‐50stk‐902561p
43. Apotek1.Freestylelansetter28G200stkNorway[cited01.03.2017].Availablefrom:https://www.apotek1.no/produkter/freestyle‐lansetter‐28g‐200stk‐951559p
44. DennorskelegeforeningNormaltariffforavtalespesialister2015‐2016[cited01.03.2017].Availablefrom:http://normaltariffen.legeforeningen.no/pdf/Normaltariff_2015.pdf
45. Helsedirektoratet.Innsatsstyrtfinansiering(ISF):VedleggA‐kostnadsvekter2016(Excel2010)Norway[cited01.03.2017].Availablefrom:https://helsedirektoratet.no/finansieringsordninger/innsatsstyrt‐finansiering‐isf‐og‐drg‐systemet/innsatsstyrt‐finansiering‐isf#regelverk‐isf‐2016
46. Statenlegemiddelverk.RefusjonsrapportLevemir[cited].Availablefrom:legemiddelverket.no
47. GodtHaab.Egenfinansiertrehabilitering[cited01.03.2017].Availablefrom:http://godthaab.no/rehabilitering/egenfinansiert‐rehabilitering/
48. Statenslegemiddelverk.RefusjonsrapportPradaxa[cited].Availablefrom:legemiddelverket.no
49. KlempM.EffektogkostnadseffektivitetavulikedialysemetoderforpasientermedalvorlignyresviktiNorge2016.
92
50. Unilabs.PrislisteforselvbetalendepasienterhosUnilabsRøntgenMajorstuenNorway[cited01.03.2017].Availablefrom:http://www.unilabs.no/Global/_NO/Prislister/Prisliste%20selvbetalende%207%20januar%202015%20Majorstuen.pdf
51. GhatnekarO,PerssonU,WillisM,OdegaardK.CosteffectivenessofBecaplermininthetreatmentofdiabeticfootulcersinfourEuropeancountries.Pharmacoeconomics2001;19(7):767‐78.
52. ClarkeP,GrayA,HolmanR.Estimatingutilityvaluesforhealthstatesoftype2diabeticpatientsusingtheEQ‐5D(UKPDS62).MedDecisMaking2002;22(4):340‐9.
53. LauridsenJT,LonborgJ,GundgaardJ,JensenHH.Diminishingmarginaldisutilityofhypoglycaemicevents:resultsfromatimetrade‐offsurveyinfivecountries.QualLifeRes2014;23(9):2645‐50.
54. Diabetesforbundet.Omdiabetes[cited01.03.2017].Availablefrom:http://diabetes.no/om‐diabetes/
55. HELFO.ATC‐koderA10AogA10B,antallpasienter,AUP,gruppe[cited01.03.2017].Availablefrom:www.helfo.no
56. HauteAutoritedeSante.FreeStyleLibreFlashGlucoseMonitoringSystem.France:NationalCommitteefortheEvaluationofMedicalDevicesandHealthTechnologies(CNEDiMTS);ReviewC;2016.
57. GøtzschePC.Whyweneedabroadperspectiveonmeta‐analysis.BMJ2000;321(7261):585.
58. AhnR,WoodbridgeA,AbrahamA,SabaS,KorensteinD,MaddenE,etal.Financialtiesofprincipalinvestigatorsandrandomizedcontrolledtrialoutcomes:crosssectionalstudy.BMJ2017;356.
59. BekelmanJE,LiY,GrossC.Scopeandimpactoffinancialconflictsofinterestinbiomedicalresearch:asystematicreview.JAMA2003;289(4):454‐65.
60. Als‐NielsenB,ChenWFGC,KjaergardLL.Associationoffundingandconclusionsinrandomizeddrugtrials:areflectionoftreatmenteffectoradverseevents?JAMA2003;290(7):921‐8.
61. PonsJ.DispositiuFlashFreeStyleLibreperalmonitoratgedelaglucemiaBarcelona:AgenciadeQualitatiAvaluacioSanitariesdeCatalunya,GeneralitatdeCatalunya;2016.
62. GobiernodeCanarias.Efectividad,seguridadycosteefectividaddelsistemaflashdemonitorizacióndeglucosaenlíquidointersticial(FreeStyleLibre®)paralaDiabetesMellitustipo1y2.Madrid:SESCS‐InformesdeEvaluaciondeTecnologiasSanitarias,MinisteriodeSanidad,ServiciosSocialeseIgualdad,,ServicioCanariodelaSalud;2016Availablefrom:http://www3.gobiernodecanarias.org/sanidad/scs/content/7f455ca1‐df18‐11e6‐83bc‐5d3a19128004/SESCS_2016_FreeStyleLibre.pdf
63. ZorginstituutNederland.Flashglucosemonitoringsysteem(FreeStyleLibre).2016.
64. BLFsdelföreningförendokrinologiochdiabetes.Riktlinjerforkontinuerligglukosmatning–continuousglucosemonitoring‐(CGM)forbarnochungdomarmedT1DMBLF’sdelföreningförendokrinologiochdiabetes;diabetesBsdfeo;2016.
65. Tandvårds‐ochläkemedelsförmånsverket.Nationellutvärderingavmedicintekniskaprodukterförenmerjämlikvård.Delrapporteringavregeringsuppdrag(S2016/07779/RS[delvis])attgörahälsoekonomiskabedömningaravmedicintekniskaprodukter.Stockholm,Sweden:TLV;2017.896/2017.Availablefrom:https://www.tlv.se/Upload/Medicinteknik/Delrapport_medicinteknik_2017.pdf
93
66. EdgeJ,AceriniC,CampbellF,Hamilton‐ShieldJ,MoudiotisC,RahmanS,etal.Analternativesensor‐basedmethodforglucosemonitoringinchildrenandyoungpeoplewithdiabetes.ArchDisChild2017.
67. CampbellF,EdgeJ,AceriniC,Hamilton‐ShieldJ,MoudiotisC,RahmanS,etal.Clinicalaccuracyevaluationoffreestylelibreflashglucosemonitoringsystemwhenusedbychildrenandyoungpeoplewithdiabetes.DiabetesTechnologyandTherapeutics2016;18:A29.
68. BaileyT,BodeBW,ChristiansenMP,KlaffLJ,AlvaS.Theperformanceandusabilityofafactory‐calibratedflashglucosemonitoringsystem.DiabetesTechnologyandTherapeutics2015;18:S18‐S9.
69. DagensDiabetes.SvarfrångeneraldirektörSofiaWallström,TLVangåendeCGM/FGMGöteborg,Sweden[citedMay2].Availablefrom:http://dagensdiabetes.info/index.php/alla‐senaste‐nyheter/2492‐svar‐fran‐generaldirektor‐sofia‐wallstrom‐tlv‐angaende‐cgm‐fgm
70. BrownP,BrunnhuberK,ChalkidouK,ChalmersI,ClarkeM,FentonM,etal.Howtoformulateresearchrecommendations.BMJ:BritishMedicalJournal2006;333(7572):804‐6.
71. NCT02821117,AbbottDiabetesCare.FreeStyleLibreGlucoseMonitoringSystemPaediatricStudy.2016.
72. RaiS,HulseA,KumarP.FeasibilityandacceptabilityofambulatoryglucoseprofileinchildrenwithType1diabetesmellitus:Apilotstudy.IndianJEndocrinolMetab2016;20(6):790‐4.
73. NCT02665455,AbbottDiabetesCare.FreeStyleLibreinPregnancyStudy.2016.74. AbererF,HajnsekM,RumplerM,ZenzS,BaumannPM,ElsayedH,etal.Evaluation
ofsubcutaneousglucosemonitoringsystemsunderroutineenvironmentalconditionsinpatientswithtype1diabetes.DiabetesObesMetab2017;19(7):1051‐5.
75. BonoraB,MaranA,CiciliotS,AvogaroA,FadiniGP.Head‐to‐headcomparisonbetweenflashandcontinuousglucosemonitoringsystemsinoutpatientswithtype1diabetes.JEndocrinolInvest2016;39(12):1391‐9.
76. DoverAR,StimsonRH,ZammittNN,GibbFW.FlashGlucoseMonitoringImprovesOutcomesinaType1DiabetesClinic.JDiabetesSciTechnol2016;26:26.
77. OlafsdottirAF,AttvallS,SandgrenU,DahlqvistS,PivodicA,SkrticS,etal.AClinicalTrialoftheAccuracyandTreatmentExperienceoftheFlashGlucoseMonitorFreeStyleLibreinAdultswithType1Diabetes.(1557‐8593(Electronic)).
78. FokkertMJ,vanDijkPR,EdensMA,AbbesS,deJongD,SlingerlandRJ,etal.PerformanceoftheFreeStyleLibreFlashglucosemonitoringsysteminpatientswithtype1and2diabetesmellitus.BMJ2017;5(1).
79. Ish‐ShalomM,WainsteinJ,RazI,MosenzonO.ImprovementinGlucoseControlinDifficult‐to‐ControlPatientsWithDiabetesUsingaNovelFlashGlucoseMonitoringDevice.JDiabetesSciTechnol2016;10(6):1412‐3.
80. JiL,GuoX,GuoL,RenQ,YuN,ZhangJ.AMulticenterEvaluationofthePerformanceandUsabilityofaNovelGlucoseMonitoringSysteminChineseAdultsWithDiabetes.JDiabetesSciTechnol2016;23:23.
81. JPRN‐UMIN000018692.Evaluationofvalidityoftheadvancedglucosemonitoring(FlashGlucoseMonitoring).2015.
82. NCT02776007,RabinMedicalCenter,Abbott.PatientsPerceptionsofUsingthe"Libre"SystemComparedWithConventionalSMBGinAdolescentsWithType1DiabetesTheLibreSatTrial.2017.
83. NCT02809365,WolfsonMedicalCenter.FreeStyleLibre‐EffectonQOLinType2DiabetesPatients.2017.
84. ACTRN12616001695493.Canwepreventrecurrentseverelowbloodglucosereactionsusinganewglucosesensingdevice?2016.
94
85. NCT02824549,JessaHospital.AnEvaluationoftheFreeStyleFlashGlucoseMonitoringSystem.2016.
86. NCT02898714,UniversitaireZiekenhuizenLeuven.FlashGlucoseMonitoringStudyforDiabetes.2016.
87. NCT02734745.AccuracyofFreestyleLibre.UniversityofPadova;2016.88. JPRN‐UMIN000023593.UseofFreeStyleLibreFlashGlucoseMonitoringSystem
andtheEffectonHypoglycaemiainPeoplewithType2DiabetesinJapan.2016.89. NCT03020264,NantesUniversityHospital.FrequencyofHypoglycemiain
PatientsWithType2DiabetesUnderInsulinTherapyOlderThan75YearsinRealLife.2017.
90. ISRCTN12543702.MaskedperformancecheckoftheAbbottFreeStyleLibreFlashGlucoseMonitoringSystem:aninterventiontrial.2015.
91. ISRCTN87654534.PerformanceCheckoftheAbbottFreeStyleLibreFlashGlucoseMonitoringSystem:amulti‐centreprospectiveopensinglearmstudy.2015.
92. HoogmaRP,HammondPJ,GomisR,KerrD,BruttomessoD,BouterKP,etal.Comparisonoftheeffectsofcontinuoussubcutaneousinsulininfusion(CSII)andNPH‐basedmultipledailyinsulininjections(MDI)onglycaemiccontrolandqualityoflife:resultsofthe5‐nationstrial.DiabetMed2006;23(2):141‐7.
93. WeinbergerM,KirkmanMS,SamsaGP,CowperPA,ShortliffeEA,SimelDL,etal.TheRelationshipbetweenGlycemicControlandHealth‐RelatedQualityofLifeinPatientswithNon‐Insulin‐DependentDiabetesMellitus.MedCare1994;32(12):1173‐81.
94. ChristiansenSC,FougnerAL,StavdahlØ,KölleK,EllingsenR,CarlsenSM.AReviewoftheCurrentChallengesAssociatedwiththeDevelopmentofanArtificialPancreasbyaDoubleSubcutaneousApproach.DiabetesTher2017:1‐18.
95. NCT02388815,AbbottDiabetesCare.AbbottSensorBasedGlucoseMonitoringSystemPaediatricStudy.2015.
96. NCT02677454,VastraGotalandRegion.EvaluationofaFlashGlucoseMonitoringSysteminAmbulatoryPatientsWithType1Diabetes.2015.
97. NCT02614768,MedicalUniversityofGraz.StandardizedEvaluationofSubcutaneousGlucoseMonitoringSystemsUnderRoutineEnvironmentalConditions.2015.
98. NTR5348.ValidationFreestylecontinuousregistrationofglucoseforthepurposeofsafeandcorrectuseintheNetherlands.NederlandsTrialRegister:http://www.trialregister.nl/trialreg/admin/rctview.asp?TC=5348;2015.
99. NCT02073058.EvaluationoftheAccuracyoftheAbbottSensorBasedInterstitialGlucoseMonitoringSystem.https://clinicaltrials.gov/ct2/show/record/NCT02073058?term=NCT02073058&cond=Diabetes+Mellitus&rank=1:ClinicalTrials.gov;2014.
95
Appendix1.Glossaryofterms
Cost‐effectivenessanalysis:aneconomicanalysisthatconvertseffectsintohealthtermsanddescribesthecostsforadditionalhealthgain.
HbA1c:thetermHbA1creferstoglycatedhaemoglobin.TheHbA1ctestmeasuresdia‐betesmanagementovertwotothreemonths.
HELFO:TheNorwegianHealthEconomicsAdministration.HELFOisresponsiblefordi‐rectpaymentstovarioushealthserviceproviders,individualreimbursementforcer‐tainmedicines,dentalservicesandhealthservicesabroad.
Health‐relatedqualityoflife:isanindividual'soragroup'sperceivedphysicalandmentalhealthovertime.
Hypocalcaemia:thisisatermthatreferstolowbloodcalciumlevel.HyperglycaemicandhypoglycaemicAUC:theareaunderthecurveistheproductofthemagnitudeanddurationofthesensormeasuredglucoselevelaboveorbelowaspecifiedcutofflevel.Highervaluesforthiscalculationindicatemorenumerous,se‐vere,orprotractedglycemicevents.
Impairedawarenessofhypoglycaemia:atermusedtodescribeasituationwherepeoplewithdiabetes,usuallytype1diabetes,arefrequentlyunabletonoticewhentheyhavelowbloodsugar.Incrementalcost‐effectivenessratio(ICER):thedifferenceinthemeancostsoftwointerventionsinthepopulationofinterestdividedbythedifferenceinthemeanout‐comesinthepopulationofinterest.Millimole:afigurethatisapartoftheInternationalSystemofUnitsandequalto1/1,000ofamole(SI).Amoleisaunitofmeasurementcommonlyusedinchemistrytodenoteamountsofatomsandmolecules.
Nephropathy:thisisatermthatreferstodamagetoordiseaseofakidney.Neuropathy:thisisatermthatreferstodiseaseordysfunctionofoneormoreperiph‐eralnerves,typicallycausingnumbnessorweakness.Retinopathy:Diabeticretinopathyisacommoncomplicationofdiabetes.Itoccurswhenhighbloodsugarlevelsdamagethecellsatthebackoftheeye(knownastheret‐ina).Ifitisn'ttreated,itcancauseblindness.
96
Type1diabetes:Diabeteswherethebodydoesnotproduceinsulin.Type2diabetes:thebodyusuallystillproducessomeinsulin,butthisisnotenoughtomeetdemandandthebody'scellsdonotproperlyrespondtotheinsulin.
97
Appendix2.Searchstrategy
Literaturesearch–FreeStyleLibreDateRun: 2017.01.18Databases: OvidMEDLINE,Embase(Ovid),CochraneLibrary;CochraneDatabaseof
SystematicReviews,OtherReviews;CentralRegisterofControlledTri‐als);TechnologyAssessments;EconomicEvaluationsDatabase,CentreforReviewsandDissemination;DatabaseofAbstractsofReviewsofEf‐fects;HealthTechnologyAssessments,PubMed.
Othersources:Googlescholar,HTAagencies,ClinicalTrials.gov,WHOInternationalClinicalTrialsRegistryPlatform.
Allresults: 1662recordsofwhich40ongoingtrials(2197includingduplicates)Searchedby: IngridHarboe,peerreviewedbyIngvildKirkeheiSearchstrategiesDatabase: CochraneLibraryAllResults: CochraneReviews(2);OtherReviews(8);Trials(175);TechnologyAs‐
sessments(11)
ID Search Hits
#1 MeSH descriptor: [Diabetes Mellitus] explode all trees 19557
#2 diabet*:ti,ab,kw
in Cochrane Reviews (Reviews and Protocols)
418
#3 diabet*
in Other Reviews, Trials, Technology Assessments and Economic Evaluations
48765
#4 #1 or #2 or #3 49246
#5 MeSH descriptor: [Blood Glucose Self-Monitoring] this term only 599
#6 ((glucose) and (monitor* or scan*)):ti,ab,kw
in Cochrane Reviews (Reviews and Protocols)
31
#7 ((glucose) and (monitor* or scan*))
in Other Reviews, Trials, Technology Assessments and Economic Evaluations
4937
#8 #5 or #6 or #7 4968
#9 (flash* or interstitial* or real-time or realtime):ti,ab,kw
in Cochrane Reviews (Reviews and Protocols)
56
#10 (flash* or interstitial* or real-time or realtime)
in Other Reviews, Trials, Technology Assessments and Economic Evaluations
7494
#11 #9 or #10 7550
98
#12 #4 and #8 and #11 196
#13 (freestyle and (flash* or libre)):ti,ab,kw i
n Cochrane Reviews (Reviews and Protocols)
0
#14 (freestyle and (flash* or libre))
in Other Reviews, Trials, Technology Assessments and Economic Evaluations
6
#15 ((guardian and (real-time or realtime)) and diabet*):ti,ab,kw
in Cochrane Reviews (Reviews and Protocols)
0
#16 ((guardian and (real-time or realtime)) and diabet*):ti,ab,kw
in Other Reviews, Trials, Technology Assessments and Economic Evaluations
9
#17 #13 or #14 or #15 or #16 15
#18 #12 or #17 196
99
Database:Embase<1974to2017January17>,EpubAheadofPrint,In‐Process&OtherNon‐IndexedCitations,OvidMEDLINE(R)DailyandOvidMEDLINE(R)<1946toPresent>
SearchStrategy:# Searches Results
1 exp Diabetes mellitus/ 1212267
2 diabet*.tw. 1292593
3 or/1-2 1524496
4 Blood Glucose Self-Monitoring/ 20461
5 blood glucose monitoring/ 21246
6 (glucose and (monitor* or scan*)).tw. 74233
7 or/4-6 90759
8 (flash* or interstitial* or real-time or realtime).tw. 727888
9 3 and 7 and 8 2547
10 (freestyle and (flash* or libre)).tw. 38
11 (guardian and (real-time or realtime) and diabet*).tw. 112
12 or/10-11 150
13 9 or 12 2551
14 remove duplicates from 13 1747
15 14 use oemezd [Embase] 1573
16 14 use ppez [MEDLINE] 174
Database:CentreforReviewsandDissemination:DARE(DatabaseforReviews
andDissemination)andHTA(HealthTechnologyAssessmentDa‐tabase)
Results: 22records Line Search Hits
1 MeSH DESCRIPTOR Diabetes Mellitus EXPLODE ALL TREES 2427
2 (diabet*) 4453
3 #1 OR #2 4459
100
4 MeSH DESCRIPTOR Blood Glucose Self-Monitoring EXPLODE ALL TREES 110
5 ((glucose and (monitor* or scan*))) 371
6 #4 OR #5 371
7 ((flash* or interstitial* or real-time or realtime)) 411
8 #3 AND #6 AND #7 22
9 ((freestyle and (flash* or libre))) 0
10 ((guardian and (real-time or realtime) and diabet*)) 1
11 #9 OR #10 1
12 #8 OR #11 22
Source:ClinicalTrials.gov
WHOICTRP(InternationalClinicalTrialsRegistryPlatform)Results:40recordsSearch:"GlucoseMonitoringSystem"ANDflash;DiabetesANDflash;freestylelibreHTAagencies(orsimilar):AHRQ(AgencyforHealthcareResearchandQuality),AHTA(AdelaideHealthTechnologyAssessment),(CADTH,QueenslandHealth,DIMDI(GermanInstituteofMedicalDocumentationandInformation)ECRI,IHE(InstituteofHealthEconomics),NICE(NationalInstituteforHealthandCareExcellence),McGillTechnologyAssessmentUnit,SBU(SwedishAgencyforHealthTechnologyAssessmentandAssessmentofSocialSer‐vices).Results:7recordsSearch:"GlucoseMonitoringSystem"ANDflash;DiabetesANDflash;freestylelibreGooglescholar:Results:2recordsSearch:Search:"GlucoseMonitoringSystem"ANDflash;DiabetesANDflash;freestylelibre
101
Appendix3.Completed(publishedandunpublished)singlearmstudies
Population Author Study Design (country)
Duration Outcomes
Type 1- Children 4 to 17yr
Edge 2017 BEAGLE NCT02388815 (66;95)
Single arm (UK)
14 days Accuracy, safety and acceptability
Type 1- Children 0 to 18yr
Rai 2016 (72)
Single arm (India)
14 days Feasibility and accepta-bility
Type 1- Children 4 to 17yr
NCT02821117 (completed) SELFY (71)
Single arm (Germany, Ireland, UK)
10 weeks Change in time in range, time in range
Type 1 or 2 Pregnant women,
NCT02665455 (completed) FLIPS (73)
Single arm (Austria, UK)
14 days Accuracy
Type 1 - adults Olafsottir 2017 NCT02677454 (77;96)
Single arm (Sweden)
10-14 days Accuracy and treatment experience
Type 1 - adults Aberer 2017 NCT02614768 (74;97)
Single arm (Austria)
12 hours 3 systems performance
Type 1 – adults Bonora 2016 (75) Ethics approval prot no 74889
Single arm (Italy)
14 days FreeStyle Libre and CGM Dexcom G4 Plati-num performance
Type 1- adults Dover 2016 (76)
Single arm (UK)
16 week Impact of FreeStyle to patients attending dia-betes clinic in a univer-sity teaching hospital
Type 1 & 2 adults Ji 2016 (80)
Single arm (China)
14 days Performance in Chinese patients
Type 1 & 2 adults Ish Shalom 2016 (79) Single arm (Israel)
12 weeks Experiences in difficult to control diabetes
Type 1 & 2 adults Fokket et al (TC5348) (78;98)
Single arm (The Netherlands)
14 days Performance
Type 1 & 2 adults Bailey 2015 NCT02073058 (68;99)
Single arm (US)
14 days Performance and usa-bility
Adults on insulin treatment
UMIN000018692 (completed) (81)
Single arm (Japan)
14 days WHO-5 questionnaire, DTSQ, safety
102
Appendix4.FreeStyleLibreregistryrecords
Population Registry number-(country)-Title Duration Outcomes
Randomized controlled trials
Type 1 Adolescents 12- 17yr
NCT02776007 (82) – (Israel) - Patients Perceptions of Using the "Libre" System Compared With Conventional SMBG in Adolescents With Type 1 Dia-betes The Libre Sat Trial Recruiting
12 weeks Treatment satisfaction, HbA1c change, average fasting blood glu-cose, sensor readings at week 12 and 24 hours, number of measure-ments on intervention and each arm, percentage of readings with 70 to 180 mg/dl, below 60mg/dl, above 240mg/dl
Type 2
Adults
NCT02809365 (83) – (Israel) - FreeStyle Libre-effect on QOL in Type 2 diabetes patients Not yet recruiting
10 weeks QoL, treatment satisfaction, HbA1c change, percentage to reach HbA1c as per physician prescrip-tion, reduction in hypoglycemic events
Adults with se-vere hy-poglycaemia
ACTRN12616001695493 (84) – (Aus-tralia)- Assessment of the utility of the Flash Libre subcutaneous glucose moni-toring system to prevent recurrent severe hypoglycaemia in patients with diabetes
Recruiting
24 weeks Recurrence or occurrence of se-vere/non-severe hypoglycaemia, time to first recurrence or occur-rence severe/non-severe hypogly-caemia, time between recruitment and first occurrence/recurrence of severe and non-severe hypogly-caemia
Single Arm/Non Randomized
Type 1
Children 4 to 18 yr
NCT02824549 (85) – (Belgium) An Evaluation of the FreeStyle Flash Glucose Monitoring System
Recruiting
24 weeks?
“0.5 yr”
Usability, skin reactions to sensor, accuracy
Type unspeci-fied
children, adults, seniors
NCT02898714 (86) – (Belgium) Flash Glucose Monitoring Study for Dia-betes (FUTURE)
Recruiting
3 years Quality of life, change in HbA1c
Type 1 -Adults
Recruiting
NCT02734745 (87) – (Italy) Accuracy of FreeStyle Libre
Recruiting
14 days Accuracy
Type 2 – ≥ 75 yrs of age
NCT03020264 (89) – (France) Frequency of hypoglycemia inpatients with type 2 diabetes under insulin ther-apy older than 75 years in real life (HY-POAGE)
4 weeks Number of confirmed or severe hy-poglycemia, events/patient/month of severe hypoglycemia, number of hospitalizations due to hypoglyce-mia, death, falls, transition to ER,
103
hypoglycemia in subroups, fear of hypoglycemia, pseudo hypoglyce-mia and unconfirmed hypoglyce-mia, nocturnal hypoglycemia, threshold blood glucose <0.54g/L, number of elderly admitted to EHPAD
Type 2 -adults UMIN000023593 (88) – (Japan) Use of FreeStyle Libre Flash Glucose Monitoring System and the Effect on Hy-poglycaemia in People with Type 2 Dia-betes in Japan Pre-initiation
12 weeks Change from baseline in time in hypoglycaemia (<3.9 mmol/L [70 mg/dL]), time in range, hyper/hypo glygaemic episodes, accuracy, var-iability measures, glucose rate change, TDD of insulyne, body weight, blood pressure, treatment satisfaction, hypoglycemia patient questionnaire, number of scans performed, user questionnaire, HCP questionnaire
Type 1 and 2 - adults
ISRCTN12543702 (90) – ( UK) Masked performance check of the Abbott FreeStyle Libre Flash glucose monitoring system. Recruiting
14 days Accuracy, precision within sensor lot
Type 1 and 2 - adults
ISRCTN87654534 (91) – (UK) Performance check of the Abbott Free-Style liver flash glucose monitoring sys-tem Recruiting
14 days Accuracy, precision within sensor lot, relationship between HbA1c levels and glycaemic variability
104
Appendix5.Includedstudiesinclusionandexclusioncriteria
InclusionCriteriaBolinder 2016
Adults 18 years or older diagnosed with type 1 DM for 5 yrs or
longer, on their current insulin regimen for at
least 3 months before study entry,
had a screening HbA1c concentration of 58mmol/mol (7.5%) or lower,
reported self monitoring of blood glucose levels on regular basis (equivalent to > 3 times/day) for 2 months or more before study entry,
considered technically capable of using the flash sensor base glucose monitoring system
Haak 2017 adults 18 years or older individuals with type 2 DM treated with
insulin for at least 6 months, on thier current regime (prandial only or
prandial and basal intensive insulin therapy or CSII therapy) for 3 months or more,
had a screening HbA1c concentration level 58-108mmol/mol (7.5-12%)
reported self monitoring regular blood glucose testing (more 10/week for at least 2 months prior to study entry)
were considered by the investigator to
be technically capable of using the flash sensor based glucose monitoring system.
Exclusioncriteria Bolinder 2016
currently diagnosed with hypoglycemia unawareness;
diabetes ketoacidosis or myocardial infraction in the preceding 6 months;
known allergy to medical grade adhesives,
used continuous glucose monitoring within the preceding 4 months,
currently using sensor augmented pump therapy,
pregnant or were planning pregnancy, or receiving oral steroid therapy for any
disorder
Haak 2017
if they had any other insulin regime to that described in inclusion;
ketoacidosis or hypersmolar-hypoglycemia state in the preciding 6 months,
if the had a total insulin >1.75 units/kg on study entry;
known allergy to medical grade adhesives
had severe hypoblycemia (requirieng third party assistance),
were pregnant or planning pregnancy, used continuous glucose monitoring
within the previous 4 months, were receiving steroid therapy for any
condition considered by the investigator
unsuitable to participate
105
Appendix 6. Results from Meta-Analysis
Outcome Type 1 DM MD[95%CI]1
Type 2 DM MD[95%CI]1
Overall MD[95%CI]1
Health Related Quality of life2 -0.10 [-0.25, 0.05] 0.00 [-0.16, 0.16] -0.05 [-0.16, 0.05]
Treatment Satisfaction3 6.20 [4.54, 7.86] 4.00 [2.32, 5.68] 5.10 [2.95, 7.26]
Pain n/a n/a
HbA1c -0.01 [-0.18, 0.16] 0.03 [-0.26, 0.32] -0.00 [-0.14, 0.14]
Glycemic Measures
Time in Range 3.9-10mmol/L3 1.20 [0.46, 1.94] 0.40 [-0.93, 1.73] 1.00 [0.32, 1.68]
<3.9mmol/L within 24 hours2 -1.24 [-1.82, -0.66] -0.40 [-0.72, -0.08] -0.60 [-0.88, -0.32]
<3.9mmol/L within 24 hours-events2 -0.37 [-0.58, -0.16] -0.15 [-0.30, 0.00] -0.23 [-0.35, -0.10]
<3.1mmol/L at night within 7 hours2 -0.35 [-0.51, -0.19] -0.10 [-0.20, -0.00] -0.22 [-0.46, 0.03]
<3.1mmol/L night within 7hr/events2 -0.11 [-0.18, -0.04] -0.07 [-0.12, -0.02] -0.09 [-0.13, -0.04]
<3.1mmol/L within 24 hours2 -0.85 [-1.24, -0.46] -0.18 [-0.35, -0.01] -0.49 [-1.15, 0.16]
<3.1mmol/L within 24 hours/events2 -0.36 [-0.53, -0.19] -0.10 [-0.19, -0.01] -0.22 [-0.47, 0.03]
>10mmol/L – time2 n/a 0.00 [-1.44, 1.44]
>13mmol/L – time2 -0.39 [-0.77, -0.01] -0.40 [-1.52, 0.72] -0.39 [-0.75, -0.03]
Moderate to severe adverse events 0.50 [0.09, 2.68]4 1.51 [0.16, 14.27]4 0.74 [0.19, 2.85]4 1MD: mean difference; [95% CI: confidence interval]; 2 lower scores are best, 3 higher scores are best; 4RR: risk ration; n/a: not assessed.
Norwegian Institute of Public Health August 2017P.O. Box 4404 NydalenNO-0403 OsloPhone: +47 21 07 70 00