Increased healthcare utilization costs following initiation of insulin treatment in type 2 diabetes: A long-term follow-up in clinical practice

ContentslistsavailableatScienceDirect

Primary

Care

Diabetes

journal homepage:http://www.elsevier.com/locate/pcd

Original

research

Increased

healthcare

utilization

costs

following

initiation

of

insulin

treatment

in

type

2

diabetes:

A

long-term

follow-up

in

clinical

practice

Almina

Kalkan

_,

_Johan

_Bodegard

_,

_Johan

_Sundström

_,

_Bodil

_Svennblad

_,

Carl

Johan

Östgren

_,

_Peter

_Nilsson

_Nilsson

_,

_Gunnar

_Johansson

_,

Mattias

Ekman

a_{AstraZenecaNordic-Baltic,Södertälje,Sweden} b_{UppsalaUniversity,Uppsala,Sweden} c_{LinköpingUniversity,Linköping,Sweden} d_{LundUniversity,Malmö,Sweden}

a

r

t

i

c

l

e

i

n

f

o

Received7July2016 Receivedinrevisedform 4November2016

Accepted6November2016 Availableonline25November2016

Keywords:

Type2diabetesmellitus Healthcareutilization Healthcarecosts Observationalstudy

a

b

s

t

r

a

c

t

Aims:Tocomparelong-termchangesinhealthcareutilizationandcostsfortype2diabetes patientsbeforeandafterinsulininitiation,aswellashealthcarecostsafterinsulinversus non-insulinanti-diabetic(NIAD)initiation.

Methods:Patientsnewlyinitiatedoninsulin(n=2823)wereidentifiedinprimaryhealthcare recordsfrom84Swedishprimarycarecenters,between1999to2009.First,healthcarecosts perpatientwereevaluatedforprimarycare,hospitalizationsandsecondaryoutpatientcare, beforeanduptosevenyearsafterinsulininitiation.Second,patientsprescribedinsulinin secondlinewerematchedtopatientsprescribedNIADinsecondline,andthehealthcare costsofthematchedgroupswerecompared.

Results:ThetotalmeanannualhealthcarecostincreasedfromD1656perpatient2years beforeinsulininitiationtoD3814sevenyearsafterinsulininitiation.Thetotalcumulative meanhealthcarecostperpatientatyear5aftersecond-linetreatmentwasD13,823inthe insulingroupcomparedtoD9989intheNIADgroup.

Conclusions: Initiationof insulin in type2 diabetes patients wasfollowed byincreased healthcarecosts.Theincreasesincostswerelargerthanthoseseeninamatchedpatient populationinitiatedonNIADtreatmentinsecond-line.

©2016TheAuthor(s).PublishedbyElsevierLtdonbehalfofPrimaryCareDiabetesEurope. ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons. org/licenses/by-nc-nd/4.0/).

∗ _{Correspondingauthorat:AstraZenecaNordic-Baltic,DepartmentofHealthEconomics,Astraallén,B674,SödertäljeSE-15185,Sweden.} E-mailaddresses:almina.kalkan@astrazeneca.com,almina.kalkan@gmail.com(A.Kalkan).

http://dx.doi.org/10.1016/j.pcd.2016.11.002

1751-9918/©2016TheAuthor(s).PublishedbyElsevierLtdonbehalfofPrimaryCareDiabetesEurope.Thisisanopenaccessarticleunder theCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1.

Introduction

Theprevalenceofdrugtreatedtype2diabetesmellitus(T2DM) in Sweden is increasing and has recently been reported to be 4.4% [1]. T2DM is a major cause of morbidity and premature mortality, primarilythrough macrovascular and microvascularcomplications[1–5].Thediseaseandits compli-cationsincreasetheuseofhealthcareserviceswithassociated increasesintotalhealthcarecosts[6–8].

GlycemiccontrolisacornerstoneinT2DMmanagement toavoiddiabetesrelatedcomplications,whereinsulinis con-sideredtobeaneffectiveHbA1cloweringintervention[9].In Sweden, witha relatively high use ofinsulin compared to otherEuropeancountries,insulinhasnowsurpassed sulpho-nylureaasthemostcommonlydispenseddruginsecond-line add-ontometformin[10,11].This“treatmentladder”isinline withtheSwedishnationalguidelines,recommendingsecond lineinsulinwhenmetforminfails,onlysubsequentlyfollowed byotheravailableglucoseloweringdrugs[12].International T2DMguidelineshoweverrecommendseveraloptionsas sec-ondline treatment,including alsomore innovativesecond linetreatment options, suchasdipeptidyl peptidase (DPP)-4inhibitors, sodiumglucosecotransporter 2inhibitorsand glucagon-likepeptide-1receptoragonists(GLP-1RA)[13,14].

Inadditiontothebenefitsonglucoselevels,insulinalso carriesanumberofunwantedsideeffectslikeweightgain, hypoglycemia,reactionsfrominjectionsandincreased treat-mentcomplexity[15–18].Furthermore,recentstudiesreport associations between insulin and increasedrisk of cancer, cardiovasculardiseaseandall-causemortality[18–20].Thus, despitethe lowdirectcost ofinsulintreatment, unwanted sideeffects and treatment complexity of insulinmay lead toincreasedlong-termhealthcarecosts[21–23].Despite sev-eralstudiesonhealthcarecostsandT2DM[24–30]thereare limiteddataonuseofhealthcareresourceassociatedwith insulin.SinceinsuliniswidelyusedinSweden,acost analy-sismightcontributeusefulevidencetotheunderstandingof theimplicationsforthehealthcaresystemoftheinsulinuse. Recently,onestudyhasreportedincreasedhealthcarecosts afterinsulininitiation,butfindingswerelimitedtoahighly selectedgroupofpatientswithnohospitalizationdata,low representativityandshortfollow-up[31].

Theaimofthisstudywastocomparelong-termchangesin healthcareutilizationandcostsbeforeandafterinsulin initi-ationinSweden.Inaddition,alsotocomparehealthcarecosts afterinsulinversusnon-insulinanti-diabetic(NIAD)initiation.

2.

Material

and

methods

2.1. Studysample

Patients diagnosedwith type2 diabetes mellitus (ICDE11) and/or prescription of any blood glucose-lowering drug (ATC A10) were identified at 84 primary-care centers in Sweden between 1 January 1999 to 31 December 2009 (www.clinicaltrials.gov; NCT:01121315). Effortwas made to ensurearepresentativeselectionofprimarycarecenters[8]. Atotalof58333patientscouldbeincluded,anddatalinkedto

theSwedishNationalPatient-,PrescribedDrug-andCauseof Deathregistriesbyusingthe uniquepersonalidentification number, mandatoryforall citizens from birthor immigra-tion. Detailsonstudy designand the dataextractionfrom primarycarerecordsandregistershavebeendescribed else-where[2,4,8,22,32].

2.2. Theinsulininitiationcohort

Thiscohortisusedtocomparelong-termchangesin health-careutilizationandcostsbeforeandafterinsulininitiationin allpatients.Thecohortwillalsobeimportantwhen assess-ingtherepresentativityofthesmallermatchedsecondline cohort,seebelow.Weidentifiedallpatients>30yearsofage initiatinginsulinafterhaving15monthswithnoinsulin pre-scriptiontobeincluded.Patientswereexcludediftheyhad noregisteredvisitorcontactintheelectronicpatientrecord withintwoyearspriortoindextreatmentstart.Anygaplarger than 15 months between prescriptions was considered as discontinuation.Inordertocontrolforothercostdriving co-morbidities,weexcludedallpatientswithhistoryofCVDand cancer atbaseline.Patients withanyhospitalvisit(in hos-pitalstay or outpatientclinicvisit)within90 dayspriorto insulininitiationwereexcluded.Patientswerefollowedfrom twoyearspriortoindexdateanduntildiscontinuation,death orendofstudyperiodfromelectronicpatientrecords.

2.3. Matchedsecondlinecohort

Inordertocomparehealthcarecostsafterinsulininitiation with healthcarecostsafterinitiation ofNIADs, wedefined two similar groups.We identifiedpatients withmetformin monotherapyforatleast2yearsandnogapsofmorethan 15 monthsbetween two prescriptions. They were indexed whentheyeitherwereprescribedsecondlineinsulin(Insulin group) or second line non-insulin antidiabetic drug (NIAD group).Toreducethelikelihoodofrescueinsulintreatment, onlypatientswithtwoprescriptionswithin15monthswere included.Anygaplargerthan15monthsbetweensecondline prescriptionswasconsideredasdiscontinuation.

2.4. Patientbaselinecharacteristics

Baselinedatawereextractedfromelectronicpatientrecords forthevariablesofsystolicanddiastolicbloodpressure; total-, lowdensity lipoprotein(LDL)and highdensitylipoprotein (HDL) cholesterol;serum triglycerides; HbA1c values; lipid-lowering-, glucose-lowering- and blood pressure-lowering drugs;andestimatedglomerularfiltrationrate(eGFR),ageand sex.DataonHbA1cisreportedinDCCT.

Diseasehistoryatbaselinewascollectedbysearchingfor diagnosescodedwithInternationalClassificationofDiseases, 9th (ICD-9)and 10th(ICD-10)revisioninprimarycare-and hospitaldata,definedinanearlierpublication[8].

2.5. Healthcareresourceuse

All patienthealthcareresourceuse (primarycare,inpatient care (hospitalizations) and/or secondary outpatient care), bothdiabetesandnon-diabetesrelated,wasconsideredand

Table1–Unitcostsappliedintheanalyses.

Costitem Unitcost(D)

Primarycare Percontact

GPvisit 148 GPhomevisit 296 GPphone contact/administrationwork 49 Nursevisit 54

Nursehomevisit 108

Nursephonecontact/patient administration

18

Otherprimarycarevisit 54

Otherprimarycarehomevisit 108

Otherprimarycarephone contact/patientadministration

18

Laboratoryvisit 57

Hospitalizations Perevent

Cardiovascular 5336 Gastrointestinal 4853 Urogenital 5067 Cancer 7081 Respiratory 5181 Endocrine 4990 Musculoskeletal 6552 Neurological 5833 Infections 5729

Allothercauses 3370

D1=8.7034SEK(2012values).

included.Dataonprimarycareusewasextractedfrom

elec-tronicpatient records and consisted ofvisits to physician,

nurse, and other primary careprofessions suchas

physio-therapist,podiatrist,laboratorytestsandadministration(e.g.,

prescriptionrenewal).

Primary care contacts were obtained by item through

theprimarycaremedicalrecords.Hospitalizationsand

sec-ondaryoutpatientcarewere extractedfrom themandatory,

SwedishNationalPatientRegister.Hospitalizationswere

clus-teredinto10diagnosis-relatedgroupsbasedonthemainICD-9

andICD-10codesassignedtoahospitalization

(cardiovascu-lar,gastrointestinal,urogenital,cancer,respiratory,endocrine,

musculoskeletal,neurological,infectiousandallothers).For

secondaryoutpatientvisitsthe 25mostfrequentdiagnoses

basedontheICD-9andICD-10wereidentified.

2.6. Costassignment

Annual healthcare costs per patient were evaluated from

the healthcareperspective forthe resourceuse inprimary

care,hospitalizationsandsecondaryoutpatientcare.Inorder

toestimate costsbefore andafter theinitiation ofinsulin,

Swedishunitcostswereappliedtothehealthcareresource

use data (Table 1). Unit costs for primary care visits were extractedfromcostdatabaseswithnationwideaveragecosts percontact.Unitcostsforhospitalizations,withthe excep-tionofcancerandothercauses,wereextractedfromthecost perpatientdatabasefromtheSwedishAssociationofLocal Authorities and Regions (KPP database). Since no uniform ICDcodeisavailableforcancer,unitcostsforcancer-related hospitalizationswereestimatedbyapplyingthemeancostper hospitalizationbasedondiagnosis-relatedgroup(DRG)cost

weights.UnitcostsforhospitalizationsthatfallsintotheICD groupofothercauseswerecalculatedbyderivingtheweighted averagecostsofthefifteenmostcommonlyobserved diag-nosesinthesample.Forsecondaryoutpatientcare,unitcosts forthe25mostfrequentICD-9andICD-10codeswereassessed usingtheKPPdatabase(AppendixA).Fortheremainingcauses in secondaryoutpatientcare,the averagecost forall diag-noseswasused.Allunitcostsareinyear2012values(average exchangerate:D1=SEK8.7034).SinceSwedenhashada neg-ligibledeflationinthegeneralpricelevelsinceyear2012,the valuesfrom2012areassumedtostillberelevant(Consumer priceindex,CPI:2015=1,2012=0.997).Costsof pharmaceuti-calsanddevicesofglucosemonitoringareexcludedfromthis studybecauseprescriptiondatawerenotavailableforthefull studyperiod.Moredetailsonthecostassessmenthavebeen describedpreviously[22].

2.7. Statisticalanalysis

Descriptivestatisticsonanaggregatedlevelwereusedto dis-playbaselinepatientcharacteristics,resourceutilization,and annualcostsperpatientinthedataset.Forcontinuous vari-ables, the mean ispresented.For categoricalvariables, the numberandproportions(percentage)ineachcategoryare pre-sented.Theperpatientmeanconsumptionofprimarycare, hospitalizationsandsecondaryoutpatientcareisreportedfor thefullstudyperiodandbyyearlyintervalsinordertoexplore patternsinresourceuseovertime.Annualcostsperpatient were estimated byapplyingthe unit costsabove, and pre-sentedintotalaswellasbyhealthcarecategory(primarycare, hospitalizationandsecondaryoutpatientcare).

TheinsulinandNIADgrouppatientswerematched1:1on meantwo-yearcostbeforeindex,HbA1c,ageandgender.

Formanypatientsthefollow-upwasshorterthanthe max-imumobservationtimeof7yearsduetodeathorcensoring attheendofthestudyperiod.Toaccountforcensoring,the partitionedmethodofBangandTsiatiswasusedfor calcu-latingthe mean annualcosts forcontactsinprimarycare, hospitalisationsandsecondaryoutpatientcare[33].

2.8. Ethicalapproval

ThestudywasapprovedbytheRegionalEthicalReviewBoard inUppsala,Sweden(Dnr.2010/077).

3.

Results

Of58,333T2DMpatients;2823patientsfulfilledthecriteriaand couldthusbeincluded.Patientsweremostfrequently initi-atedonmediumlongacting(NPH)followedbymixedmedium long-,longactingandshortactinginsulin.Patientswere fol-lowedupto7years(mean3.8yearsandtotalnumberofpatient years10,727).Meanageofpatientsatinitiationofinsulinwas 61yearsand42%werefemale.MeanpatientHbA1clevelwas 8.7%.FurtherbaselinecharacteristicsarereportedinTable2.

Table2–Baselinecharacteristics.

Allpatientsstartedwithinsulin Beforematching Matchedpatients(1:1)

Insulin NIAD Insulin NIAD

Numberofpatients,N 2823 511 1859 432 432

Age,years, 60.8(11.1) 59.4(11.0) 59.3(11.5) 59.5(10.1) 59.5(10.1)

Sex,male,n(%) 1645(58.3) 279(54.6) 1028(55.3) 243(56.2) 253(58.6)

Followup,years 3.8(2.6) 4.2(3.1) 3.3(2.6) 4.0(3.0) 3.2(2.6)

Timeonmetforminmonotherapy(days) n/a 787.7(752.2) 831.4(727.6) 795.1(760.0) 797.9(696.0)

Insulintype,n(%)

Shortacting 90(3.2) 15(3.0) 14(3.3)

Mediumacting 1558(55.2) 282(57.1) 243(57.9)

Mixedmedium/longacting 620(22.0) 92(18.6) 71(16.9)

Longacting 554(19.6) 105(21.3) 92(21.9) Cardiovasculardisease,n(%) 0 0(0) 0(0) 0(0) 0(0) Myocardialinfarction 0 0(0) 0(0) 0(0) 0(0) Stroke 0 0(0) 0(0) 0(0) 0(0) Heartfailure 0 0(0) 0(0) 0(0) 0(0) Atrialfibrillation 0 0(0) 0(0) 0(0) 0(0) Otherdisease,n(%) 0 0(0) 0(0) 0(0) 0(0) Cancer 0 0(0) 0(0) 0(0) 0(0) Kidneydisease 0 0(0) 0(0) 0(0) 0(0) Laboratorymeasurements HbA1c,% 8.7(2.8) 8.8(2.9) 8.2(2.6) 8.6(2.6) 8.6(2.6) Glucose,mmol/l 10.5(3.7) 11.1(4.1) 9.8(3.0) 10.9(3.9) 11.1(3.2) BMI,kg/m2 _{29.9(5.5) 31.5(5.2) 31.9(5.7) 31.4(5.0) 32.1(5.7)}

Systolicbloodpressure,mmHg 142.9(18.7) 141.5(17.5) 142.3(17.6) 141.6(17.3) 143.3(17.3)

Diastolicbloodpressure,mmHg 80.6(9.8) 81.6(9.2) 81.9(9.8) 81.5(9.1) 82.5(9.8)

Totalcholesterol,mmol/l 5.1(1.1) 5.0(1.2) 5.2(1.1) 5.0(1.2) 5.3(1.2)

HDLcholesterol,mmol/l 1.3(0.7) 1.4(0.8) 1.3(0.7) 1.4(0.8) 1.2(0.5)

LDLcholesterol,mmol/l 2.9(0.9) 2.9(0.9) 3.0(0.9) 2.9(0.9) 3.1(0.9)

Triglycerides,mmol/l 2.2(1.8) 2.2(1.8) 2.4(1.6) 2.2(1.8) 2.4(1.8)

Creatinine,␮mol/l 78.6(21.6) 74.7(20.0) 76.3(19.9) 74.7(19.4) 76.0(17.8)

EstimatedGFR,ml/min 84.0(20.3) 87.9(18.1) 86.8(19.4) 88.1(17.9) 87.1(18.8)

Insulin/add-ontreatmentinitiated,n(%)

1999 178(6.3) 21(4.1) 70(3.8) 17(3.9) 21(4.9) 2000 206(7.3) 27(5.3) 100(5.4) 17(3.9) 29(6.7) 2001 206(7.3) 32(6.3) 99(5.3) 24(5.6) 25(5.8) 2002 201(7.1) 36(7.0) 120(6.5) 28(6.5) 26(6.0) 2003 207(7.3) 36(7.0) 171(9.2) 33(7.6) 32(7.4) 2004 237(8.4) 36(7.0) 153(8.2) 34(7.9) 27(6.2) 2005 264(9.4) 40(7.8) 186(10.0) 37(8.6) 46(10.6) 2006 263(9.3) 46(9.0) 204(11.0) 36(8.3) 42(9.7) 2007 323(11.4) 74(14.5) 213(11.5) 58(13.4) 59(13.7) 2008 253(12.5) 76(14.9) 272(14.6) 65(15.0) 59(13.7) 2009 385(13.6) 87(17.0) 272(14.6) 83(19.2) 66(15.3)

Allnumbersinparenthesisarestandarddeviationifnotstatedotherwise.HDL,highdensitylipoprotein,LDL,lowdensitylipoprotein,GFR, glomerularfiltrationrate.

3.1. Healthcarecostsofpatientsinitiatedoninsulin

ThetotalmeanannualhealthcarecostsincreasedfromD1655

perpatient2yearsbeforeinsulininitiationtoD3814seven

yearsafterinsulininitiation(Fig.1).Almosthalfofthetotal

increase in mean annual healthcare cost occurred already inyear1,afterwhichthecostscontinuedincreasing gradu-ally.Throughoutthestudyperiod,primarycarerepresentsthe largestportionofthetotalhealthcarecosts(43–59%),closely followedbyhospitalisation(31–40%).Primarycarecostsare especiallyelevatedoneyearafterinsulininitiation,while hos-pitalisationcostsincreasesteadily overtime. Thecostsfor secondarycareoutpatientvisitskeeparelativelysmallshare (10–16%)oftotalcoststhroughoutthestudyperiod.Meanper

patientcumulativehealthcarecostswereD14,211at5years and D21,334at7years ofobservation. Primary carehad a highercumulativecostcomparedwithhospitalcare,D10,001 andD8188atyear7,respectively.

3.2. Healthcarecostsaftersecondlineinsulinversus NIAD

Overall,511patientsfulfilledtheinclusioncriteriaofhaving metforminmonotherapyforatleast2yearswithnogapsof morethan15monthsbetweentwoprescriptions,andbeing initiated on second line insulin or NIAD. The unmatched patientswhoinitiatedinsulincomparedtoNIADhadslightly higherlevelsofHbA1candbloodglucosebutweresimilarin

Fig.1–Meanannualperpatienthealthcarecosts2yearspriortoinsulinstartandtheyearsthereafter.

allotherrespects.Thematchedgroupssharedsimilar charac-teristicscomparedtothelargergroupofallpatientsinitiating insulin,hencetherepresentativityseemstobeappropriate.

Thetotalcumulativemeanhealthcarecostperpatientat year5aftersecond-linetreatmentwithinsulinwasD13,823, with hospitalizations as the main cost driverrepresenting 42–51%oftotalcumulativecosts, followedbyprimary care visits(33–46%oftotalcost)andoutpatientvisits(12–16%of total cost).The totalcumulative mean healthcarecost per patientatyear5aftersecond-linetreatmentwithNIADwas D9989,withthesamedistributionincostcategoriesasforthe insulingroup(Fig.2).Healthcarecostsaftersecondlineinsulin initiationexhibit a12-foldincrease comparedtotwoyears beforeinsulininitiation,whilethehealthcarecostsafter sec-ondlinewithNIADexhibitan8-foldincrease.Thedifference incumulativehealthcarecostsbetweensecondline insulin andsecondlineNIADappearedalreadyinyear1and contin-uedtoincreaseduringfollow-up(Fig.3).Hencetheincrease in healthcare costs associated with insulin initiation is considerablyhigherthattheincreaseseenafterNIAD initi-ationinsecond-line.

4.

Discussion

Basedon datafromalargeSwedish observationalstudy of T2DMpatients,thisstudydemonstratestwoimportant find-ings; the increased healthcare use and costs after insulin initiation;and significantlyhighercumulative costsamong

patientsinitiatedonsecondlineinsulincomparedtopatients initiatedonsecondlinenon-insulinanti-diabetics(NIADs).

The insulin initiation resulted in a sharp rise of total healthcarecostswithinthefirstyear,andwasfollowedby con-tinuouslyincreasedtotalhealthcarecostsupto7years.Mean annualhealthcarecostsperpatientincreased230%overthe 7years.ArecentSwedishstudyhasalsoreportedincreased healthcarecosts 1year afterinsulininitiation,althoughat somewhat different rates than those reportedhere [31]. In thestudybyBexeliusetal.,thecostsincreasedfromD1980 the yearbeforeinsulininitiationtoD3637alreadythe year after (2012 values, 1D=8.7034SEK). One possible explana-tionforthisdiscrepancyisthatincontrasttoBexeliusetal., using data from 100 patients in 1 Swedish county coun-cil, this study included geographically dispersed(rural and urban)primarycarecentersaswellasthenationwideSwedish NationalPatient-,PrescribedDrug-andCauseofDeath reg-istries. Another possible explanation for the difference is thestricterinclusioncriteriainthepresentstudy,excluding patientswithknownCVDorcancer,orpatientswhowere hos-pitalizedwithin90dayspriortoinsulininitiation,tocontrol for expensiveco-morbidities inorder toget moreaccurate estimates ofthechangesincosts directlylinkedtoinsulin initiation.

Baseline characteristics ofthe unmatched insulingroup show that these patients have higher HbA1c and receive slightlyearlieradd-ontreatmentthandoNIADpatients.This couldimplyamoreprogresseddiseaseorlaterdiagnosis set-tingintheinsulingroup.Apreviousstudyhasdemonstrated therelationshipbetweencostsandHbA1clevel,wherea

1%-Fig.2–Cumulativemeancostperpatientpriorandafteradd-ontreatmentinpatientstreatedwithinsulin(upperpane)and patientstreatedwithNIAD(lowerpane).

pointincreaseinHbA1cled,onaverage,toa4.4%increasein diabetes-relatedmedicalcostsfortype2diabetes[34]. This relationshipcouldhelpexplainthe costincreaseovertime asthediseaseprogresses.Inthepresentstudy,onmatching theinsulinpatientsinHbA1candotherimportantvariablesto patientsreceivingNIAD-treatmentinsecondline,theinsulin initiation was followed by a higher use of healthcare ser-vices and a larger increase in total healthcare costs, than non-insulintreatment.Toourknowledge,thepresentstudy

isthefirsttocomparehealthcarecostsofpatientsinitiated onsecondlineinsulintoamatchedgroupofpatients initiat-ingsecondlinewithanotherNIAD.Thedifferenceinhealth carecostswasvisiblealreadyafterthefirstyearand contin-uedtogrowduring5yearobservationperiod.Atyear5,insulin patientshadincurredD3800moreincumulativehealthcare costs perpatient than non-insulinpatients. Thedifference incostsisstriking,especiallyintheviewthatinsulinisput

0% 200% 400% 600% 800% 1000% 1200%

-2 -1 0 1 year 2 year 3 year 4 year 5 year

Insulin NIAD

Fig.3–Increaseincumulativemeancostperpatientyear 1–5versusdateofadd-ontreatmentinpatientstreated withinsulin(redfullline)andpatientstreatedwithNIAD (bluedashedline).(Forinterpretationofthereferencesto colorinthisfigurelegend,thereaderisreferredtotheweb versionofthisarticle.)

forwardforitsallegedlowcostcomparedtonon-insulin treat-ments[13].

The increase in primary care contacts is of particular interestasitindicatesaneedforamoreintensefollowup afterinsulininitiation.InSweden,interactionwithdiabetes teamsafterinsulininitiationiswellestablishedand recom-mendedinthenationalguidelines,whichcouldhelpexplain theincreaseinprimarycarecosts. Inaddition, hospitaliza-tions increased substantially in the first year after insulin initiation,comparedtopatientsprescribednon-insulin treat-ment.Theobserveddifferenceinhospitalizationcostsmaybe mainlyrelatedtotheadverseeffectsofinsulin,particularly hypoglycemia,thatrequireadditionalhealthcare.Despitethe known benefitswithearly insulin initiation [35–37], it also carriesanumberofunwantedsideeffects[11,15–17]. Hypo-glycemicresponses,suchassympathoadrenergicactivation, aresuspectedtoactasacausalpathwaybetweeninsulin treat-mentandriskofCVD,mortality,andpotentialseriouscardiac arrhythmias[38].

Furthermore,theincreasedhospitalizationmightbelinked toweightgain,whichisawell-knownside-effectofinsulin treatment.Severalstudieshaveshownthatoverweightisa majorcontributortotheincreasedriskforCVmorbidityand mortality in T2DM patients [8,32,39,40]. In addition, treat-ment induced weight gain has been reported as having a negativeeffectonadherencetodiabeticmedications[41]. Con-sequently,severalstudieshaveshownthatweightchangesin T2DM-patientshadeffectonhealthcarecostsduetoincreased healthcareconsumption[42–44].Forinstance,everyunitgain inBMIwasfoundtobeassociatedwith20%increaseincosts among patients who increased their BMI over 12 months (changeinBMI>0)[44].

PreviousresearchofhealthcarecostsforT2DMinSweden has evaluatedcostsfor1year atatime, andoften in lim-itedcohorts[24–31].Thisstudylongitudinallyfollowspatients’ healthcareuseandcostdevelopmentafterinitiationofinsulin inalargepatientcohort(n=2823).Patientswithdata2years priortoinsulininitiationweredeliberativelychosentodepict thechangesincostsafterinsulininitiation.Byusingalarge data material from 84 primary care centers from different partsofSweden, selectionbiashasbeenreduced.However, this study alsohas limitations.One isthe relatively small number of patients available for the full 7 year observa-tion period.Therepresentativenessofthecohortsincluded couldbequestioned,supportingtheimportancetocompare thelargerinsulininitiatingcohorttothesecond-linecohort match.Thematchedcohortwasverysimilartoallpatients whowereinitiatedoninsulin,andthe5-yearcumulativemean healthcarecostsweresimilarinbothinsulingroups(D13,823 versusD14,211),supportinghighinternalrepresentativeness. Theexclusionofpatientswho havepreviousexperienceof CVDorcancer,orwhowerehospitalizedwithin90daysprior toinsulininitiation,wasalsoperformedtominimizetherisk forconfounding.Anotherpotentialdrawbackaretheincluded costitems.Healthcareutilizationislimitedtoprimarycare visits, hospitalizationsandsecondaryoutpatientvisits, and doesnotaccountforotheremergencycareoroutpatientcare. However,themaintreatmentofT2DMinSwedenisprovided bythe includeditems, whythisstudy givesawell-covered estimateofthecoststhatfollowinsulininitiation.

5.

Conclusions

Initiation of insulin in type 2 diabetes patients were fol-lowedbyincreasedcostsinprimary,secondaryoutpatientand hospitalcare,largerthanthoseseeninamatchedpatient pop-ulationinitiatedonnon-insulintreatmentinsecond-line.To fullyanalyzetheimplicationsofinsulinandNIADtreatment, acost-effectivenessanalysisisrequired.

Conflict

of

Interest

AK,JB,andMEholdfull-timepositionsatAstraZeneca.JS,BS, CJÖ, PN and GJhave receivedcompensationfortheir work fromAstraZeneca.

Acknowledgement

ThisstudywassponsoredbyAstraZeneca.

Appendix

A.

Itemcostsusedtocalculatemeanandmediancostsper indi-vidualforoutpatientvisits.Costreference:KPPDatabase.

ICD10 EUR/admission Year Comments 297.58 2013 E11 190.16 2013 E10 290.81 2013 H25 167.52 2013 Z49 447.76 2013 Z09 237.84 2013 E14 295.40 2013 Z13 134.77 2013 H40 118.23 2013 I20 381.46 2013 C61 383.76 2013 I48 406.51 2013 Z01 183.38 2013 M17 290.23 2013 I25 312.06 2013 H35 305.97 2013 R10 341.71 2013 Z51 416.85 2013 Z08 478.89 2013 L40 277.71 2013 I70 316.77 2013 R07 392.95 2013 M79 410.30 2013 Z96 250.36 2013 M05 298.16 2013

otherICDcodes 310.91 2013 meanforall diagnosis

r

e

f

e

r

e

n

c

e

s

[1] A.Norhammar,J.Bodegard,T.Nystrom,etal.,Incidence, prevalence,andmortalityoftype2diabetesrequiring glucose-loweringtreatment,andassociatedrisksof cardiovascularcomplications:anationwidestudyin Sweden,2006–2013,Diabetologia5(2016)(Epubaheadof print).

[2] J.Sundstrom,R.Sheikhi,C.J.Ostgren,etal.,Bloodpressure levelsandriskofcardiovasculareventsandmortalityin type-2diabetes:cohortstudyof34009primarycare patients,J.Hypertens.31(8)(2013)1603–1610.

[3] ADVANCECollaborativeGroup,A.Patel,S.MacMahon,J. Chalmers,etal.,Intensivebloodglucosecontroland vascularoutcomesinpatientswithtype2diabetes,N.Engl. J.Med.358(24)(2008)2560–2572.

[4] C.J.Östgren,J.Sundström,B.Svennblad,etal.,Associations ofHbA1candeducationallevelwithriskofcardiovascular eventsin32,871drug-treatedpatientswithtype2diabetes: acohortstudyinprimarycare,Diabet.Med.30(5)(2013) e170–7.

[5] R.Williams,L.VanGaal,C.Lucioni,C.-A.Board,Assessing theimpactofcomplicationsonthecostsoftypeIIdiabetes, Diabetologia45(7)(2002)S13–7.

[6] P.E.Wandell,Qualityoflifeofpatientswithdiabetes mellitus:anoverviewofresearchinprimaryhealthcarein theNordiccountries,Scand.J.Prim.HealthCare23(2)(2005) 68–74.

[7] P.Kanavos,S.vandenAardweg,W.Schurer,Diabetes Expenditure,BurdenofDiseaseandManagementin5EU

Countries,TheLondonSchoolofEconomicsandPolitical Science(LSE),London,UK,2012.

[8] J.Bodegard,J.Sundstrom,B.Svennblad,etal.,Changesin bodymassindexfollowingnewlydiagnosedtype2diabetes andriskofcardiovascularmortality:acohortstudyof8486 primary-carepatients,DiabetesMetab.39(4)(2013)306–313. [9] UKProspectiveDiabetesStudy(UKPDS)Group,Intensive

blood-glucosecontrolwithsulphonylureasorinsulin comparedwithconventionaltreatmentandriskof complicationsinpatientswithtype2diabetes(UKPDS33), Lancet352(9131)(1998)837–853.

[10] J.Bodegård,D.Nathanson,T.Nyström,etal.,Second-line treatmentwithsulphonylureacomparedtoDPP4inhibitors demonstratedassociationswithearliertreatment

intensificationwithinsulin,in:PosterattheEuropean AssociationfortheStudyofDiabetes(EASD)Annual Meeting,Stockholm,Sweden,2015,Availablefrom: http://www.easdvirtualmeeting.org/resources/second-line- treatment-with-sulfonylurea-compared-to-dpp4-inhibitors- demonstrated-associations-with-earlier-treatment-intensification-with-insulin-3.

[11] E.Heintjes,J.Overbeek,P.Blin,etal.,Type2diabetes treatmentpatternsacrossEurope[Poster],ValueHealth17 (2014)A323–A686,Availablefrom:

http://www.harmo.nl/uloads/df/Posters/T2DM%20treatment% 20atterns%20across%20EuroeISPOR201420141030.pdf. [12] Socialstyrelsen,Nationellariktlinjerfördiabetesvård.Stöd

förstyrningochledning,NationalGuidelinesfordiabetes careinSweden,SocialstyrelsenStockholm(2015),2015-2-3. [13] S.E.Inzucchi,R.M.Bergenstal,J.B.Buse,etal.,Management

ofhyperglycemiaintype2diabetes,2015:apatient-centered approachupdatetoapositionstatementoftheAmerican DiabetesAssociationandtheEuropeanAssociationforthe StudyofDiabetes,DiabetesCare38(2015)140–149. [14] NationalInstituteforHealthandCareExcellence(NICE),

AlgorithmforBloodGlucoseLoweringTherapyinAdults withType2Diabetes,NICEguidelineNG28,2015.Available from:

https://www.nice.org.uk/guidance/ng28/resources/algorithm-for-blood-glucose-lowering-therapy-in-adults-with-type-2-diabetes -2185604173.

[15] D.Russell-Jones,R.Khan,Insulin-associatedweightgainin diabetes—causes,effectsandcopingstrategies,Diabetes Obes.Metab.9(6)(2007)799–812.

[16] B.T.Srinivasan,J.Jarvis,K.Khunti,M.J.Davies,Recent advancesinthemanagementoftype2diabetesmellitus:a review,Postgrad.Med.J.84(996)(2008)524–531.

[17] T.M.Davis,R.M.Clifford,W.A.Davis,Effectofinsulintherapy onqualityoflifeinType2diabetesmellitus:theFremantle DiabetesStudy,DiabetesRes.Clin.Pract.52(1)(2001)63–71. [18] S.E.Holden,S.Jenkins-Jones,C.L.Morgan,etal.,

Glucose-loweringwithexogenousinsulinmonotherapyin type2diabetes:doseassociationwithall-causemortality, cardiovasculareventsandcancer,DiabetesObes.Metab.17 (4)(2015)350–362.

[19] R.M.Stoekenbroek,K.L.Rensing,S.J.BernelotMoens,etal., Highdailyinsulinexposureinpatientswithtype2diabetes isassociatedwithincreasedriskofcardiovascularevents, Atherosclerosis240(2)(2015)318–323.

[20] C.L.Roumie,R.A.Greevy,C.G.Grijalva,etal.,Association betweenintensificationofmetformintreatmentwith insulinvssulfonylureasandcardiovasculareventsand all-causemortalityamongpatientswithdiabetes,JAMA311 (22)(2014)2288–2296.

[21] AmericanDiabetesAssociation,Economiccostsofdiabetes intheU.S.in2007,DiabetesCare31(3)(2008)596–615. [22] U.Sabale,J.Bodegard,J.Sundstrom,etal.,Healthcare

diabetesinSweden:afollow-upof38,956patientsina clinicalpracticesetting,Prim.CareDiabetes9(5)(2015) 330–337.

[23] B.Berger,G.Stenstrom,G.Sundkvist,Incidence,prevalence, andmortalityofdiabetesinalargepopulation:areport fromtheSkaraborgDiabetesRegistry,DiabetesCare22(5) (1999)773–778.

[24] DiabetesNationalDataUS:NationalCenterforChronic DiseasePreventionandHealthPromotion,Divisionof DiabetesTranslation.NationalDiabetesReport,2014. Availablefrom:

http://www.cdc.gov/diabetes/pubs/statsreport14/national -diabetes-report-web.pdf.

[25] J.B.Brown,G.A.Nichols,H.S.Glauber,A.W.Bakst,Type2 diabetes:incrementalmedicalcarecostsduringthefirst8 yearsafterdiagnosis,DiabetesCare22(7)(1999)1116–1124. [26] F.Henriksson,C.D.Agardh,C.Berne,etal.,Directmedical

costsforpatientswithtype2diabetesinSweden,J.Intern. Med.248(5)(2000)387–396.

[27] B.Jonsson,C.A.Board,RevealingthecostoftypeIIdiabetes inEurope,Diabetologia45(7)(2002)S5–12.

[28] M.Brandle,H.Zhou,B.R.Smith,etal.,Thedirectmedical costoftype2diabetes,DiabetesCare26(8)(2003)2300–2304. [29] A.Ringborg,M.Martinell,J.Stålhammar,etal.,Resourceuse

andcostsoftype2diabetesinSweden—estimatesfrom population-basedregisterdata,Int.J.Clin.Pract.62(5)(2008) 708–716.

[30] K.Bolin,C.Gip,A.C.Mörk,B.Lindgren,Diabetes,healthcare costandlossofproductivityinSweden1987and2005—a register-basedapproach,Diabet.Med.26(September(9)) (2009)928–934.

[31] C.Bexelius,J.Lundberg,X.Wang,etal.,Annualmedical costsofSwedishpatientswithtype2diabetesbeforeand afterinsulininitiation,DiabetesTher.4(December(2))(2013) 363–374.

[32] I.Grundvold,J.Bodegard,P.M.Nilsson,etal.,Bodyweight andriskofatrialfibrillationin7,169patientswithnewly diagnosedtype2diabetes;anobservationalstudy, Cardiovasc.Diabetol.14(2015)5.

[33] H.Glick,J.Doshi,A.Sonnad,D.Polsky,AnalyzingCensored Cost.EconomicEvaluatuationinClinicalTrial,Oxford UniversityPress,2010,pp.115–131.

[34] M.Aagren,W.Luo,Associationbetweenglycemiccontrol andshort-termhealthcarecostsamongcommercially insureddiabetespatientsintheUnitedStates,J.Med.Econ. 14(1)(2011)108–114.

[35] M.Alvarsson,G.Sundkvist,I.Lager,etal.,Effectsofinsulin vsglibenclamideinrecentlydiagnosedpatientswithtype2 diabetes:a4-yearfollow-up,DiabetesObes.Metab.10(5) (2008)421–429.

[36] L.B.Harrison,B.Adams-Huet,P.Raskin,I.Lingvay,Beta-cell functionpreservationafter3.5yearsofintensivediabetes therapy,DiabetesCare35(7)(2012)1406–1412.

[37] J.Weng,Y.Li,W.Xu,etal.,Effectofintensiveinsulintherapy onbeta-cellfunctionandglycaemiccontrolinpatientswith newlydiagnosedtype2diabetes:amulticentrerandomised parallel-grouptrial,Lancet371(9626)(2008)1753–1760. [38] S.Zoungas,A.Patel,J.Chalmers,etal.,Severehypoglycemia

andrisksofvasculareventsanddeath,N.Engl.J.Med.363 (15)(2010)1410–1418.

[39] K.Eeg-Olofsson,J.Cederholm,P.M.Nilsson,etal.,Riskof cardiovasculardiseaseandmortalityinoverweightand obesepatientswithtype2diabetes:anobservationalstudy in13,087patients,Diabetologia52(January(1))(2009)65–73. [40] J.Klenk,K.Rapp,H.Ulmer,etal.,Changesofbodymass

indexinrelationtomortality:resultsofacohortof42,099 adults,PLoSOne9(January(1))(2014)e84817.

[41] LookAHEADResearchGroup,R.R.Wing,Long-termeffects ofalifestyleinterventiononweightandcardiovascularrisk factorsinindividualswithtype2diabetesmellitus: four-yearresultsoftheLookAHEADtrial,Arch.Intern.Med. 170(September(17))(2010)1566–1575.

[42] J.Mukherjee,C.Sternhufvud,N.Smith,etal.,Association betweenweightchange,clinicaloutcomes,andhealthcare costsinpatientswithtype2diabetes,J.Manag.CareSpec. Pharm.22(5)(2016)449–466.

[43] U.Sabale,J.Bodegård,J.Sundströmetal.Changesinbody massindexovertimeandtheirassociationwithhealthcare costsamongnewly-diagnosedtype-2diabetespatientsin Sweden,forthcoming.

[44] TheECOBIMStudy,T.Dilla,A.Valladares,C.Nicolay,etal., Healthcarecostsassociatedwithchangeinbodymassindex inpatientswithtype2diabetesmellitusinSpain,Appl. HealthEcon.HealthPolicy10(6)(2012)417–430.