Pneumococcal vaccination in older adults in the era of childhood vaccination: Public health insights from a Norwegian statistical prediction study

(1)

ContentslistsavailableatScienceDirect

Epidemics

jo u rn al h om ep age : w w w . e l s e v i e r . c o m / l o c a t e / e p i d e m i c s

Pneumococcal vaccination in older adults in the era of childhood vaccination: Public health insights from a Norwegian statistical prediction study

Anneke Steens

^a,b,∗

, Didrik F. Vestrheim

^a

, Birgitte Freiesleben de Blasio

^a,c

aDivisionofInfectiousDiseaseControl,NorwegianInstituteofPublicHealth,Norway

bEuropeanProgrammeforInterventionEpidemiologyTraining(EPIET),EuropeanCentreforDiseasePreventionandControl(ECDC),Sweden

cOsloCenterforBiostatisticsandEpidemiology,DepartmentofBiostatistics,InstituteofBasicMedicalSciences,UniversityofOslo,Norway

a r t i c l e i n f o

Articlehistory:

Received30September2014

Receivedinrevisedform8January2015 Accepted12January2015

Availableonline19January2015

Keywords:

Streptococcuspneumoniae Pneumococcalvaccines Numberneededtovaccinate Time-seriesanalysis Elderly

a b s t r a c t

Twodifferentvaccines, a23-valentpolysaccharidevaccine(PPV23) anda13-valentconjugatevac- cine(PCV13),areavailableforpreventionofinvasivepneumococcaldisease(IPD)inthepopulation aged65yearsandolder(65+).TheIPDepidemiologyinthe65+isundergoingchangeduetoindi- recteffectsofchildhoodimmunisation.Vaccinerecommendationsforthe65+musttakeintoaccount thesetrendsinepidemiology.Wethereforeexploredthepreventivepotentialofvaccinationstrate- giestopreventIPDinthe65+,includingPPV23,PCV13orPCV13+PPV23in2014–2019.Quasi-Poisson regression models were ﬁttedto 2004–2014 population-wide surveillance data and used topre- dictincidences forvaccine-typeand non-vaccinetype IPD. Wedeterminedthe numberofpeople neededtobevaccinatedtopreventonecaseperseason(NNV)foreachstrategyandestimatedthe publichealth impacton the IPDcase counts from increasing thevaccine uptake to28–45%. Our resultsindicatethatPCV13-IPDwilldecreaseby71%from58(95%predictioninterval55–61)cases in2014/15to17 (6–52)in 2018/19andPPV23-IPD by32%from 168(162–175)to115 (49–313) cases.TheNNV willincreaseovertimeforallstrategiesbecause ofadecreasingvaccine-typeIPD incidence.In2018/19,thePCV13-NNVwillbe5.3timeshigherthanthePPV23-NNV.Increasingthe vaccineuptake willleadtoalarger public healthimpactforall scenarios.Combining PCV13 and PPV23is mosteffective, butthe additionaleffect ofPCV13 willdecrease and isonlymarginal in 2018/19.OurstudydemonstratestheimportanceofincreasingPPV23uptakeandofdevelopingvaccines thatconferbroaderimmunity.

1. Introduction

Streptococcuspneumoniae(pneumococci)arepartofournormal nasopharyngealﬂora,butcancauseseverediseasesuchasinvasive pneumococcaldisease(IPD;e.g.meningitis,febrilebacteraemia).

Abbreviations:IPD,invasivepneumococcaldisease;PCV,pneumococcalconjugatevaccine;PCV7,7-valentpneumococcalconjugatevaccine;PCV13,13-valent pneumococcalconjugatevaccine;PCV13-7,sixadditionalserotypesthatarein PCV13butnotinPCV7;PPV2323,valentpolysaccharidevaccine;PPV23-11,Twelve serotypesthatareinPPV23butnotinPCV13;VE,vaccineeffectiveness;NVT,Non- vaccineserotypes;NIPH,NorwegianInstituteofPublicHealth;MSIS,Norwegian SurveillanceSystemforCommunicableDiseases;95%PI,95%Predictionintervals;

NNV,Numberneededtovaccinate;PHIs,Publichealthimpactperfutureseason.

∗Correspondingauthorat:DivisionofInfectiousDiseaseControl,NorwegianInsti- tuteofPublicHealth,P.O.Box4404Nydalen,NO-0403OSLO,NORWAY.

Tel.:+4721076780;fax:+4721076513.

E-mailaddress:[email protected](A.Steens).

Olderadultsaged65yearsandolder(furthercalled65+)areamong themostvulnerablepopulationforIPD.InmostWesternsocieties, pneumococcalvaccinationisthereforerecommendedforthe65+.

In Norway,the23-valent pneumococcalpolysaccharidevaccine (PPV23)hasbeenrecommendedtothe65+ since1996,though uptakeisestimatedtobeonlyabout15–30%(unpublisheddata NIPH).TheeffectivenessofPPV23topreventpneumococcaldis- easeinolderadultsremainssubjectofcontroversy(Shapiroetal., 1991;Moberleyetal.,2013;Hussetal.,2009;Jacksonetal.,2003).

Recently,a13-valentpneumococcalconjugatevaccine(PCV13) waslicencedintheEUforuseinallagegroups(EuropeanMedicines Agency,2013),andatpresentbothPPV23andPCV13areavailable forpreventionofIPDinthe65+.TwelveofthePCV13serotypes arealsoincludedinPPV23.PCV13likelyprovidesbetterprotec- tionagainstpneumococcaldiseaseduetodifferentimmunogenic properties (Bonten et al., 2014). PCV13 has been used in the Norwegian childhood immunisation programme since 2011. In ordertomakeaninformedchoicebetweenPPV23andPCV13for http://dx.doi.org/10.1016/j.epidem.2015.01.001

(2)

the65+population,evaluationofthepreventivepotentialofthe vaccinesisneeded.

Thepotentialofavaccinetopreventdiseasedependsonthe incidenceofdiseasecausedbyserotypesthatarecoveredbythe vaccine,the vaccineeffectiveness (VE) and thevaccine uptake.

LikeinothersettingswherePCVhasbeenimplementedinchild- hoodimmunisationprogrammes(Lexauetal.,2005;Milleretal., 2011;Harboeetal.,2014),theepidemiologyofIPDinNorwayhas substantiallychanged,bothbydirectprotectionofimmunisedchil- dren, and byindirectprotectiondue todecreased transmission ofvaccine serotypepneumococci tonon-vaccinatedagegroups (Steensetal.,2013).Simultaneously,theincidenceofIPDcaused bynon-vaccineserotypes(NVT)hasslightlyincreasedasaresult of serotype replacement (Weinberger et al., 2011;Hicks et al., 2007;Milleretal.,2011;Steensetal.,2013).Furtherchangesinthe IPDepidemiologyareexpectedforthenearfuture,asitwilltake someyears beforetheserotypedistributionhasstabilised after switchingtoPCV13(Hanageetal.,2010).Suchfuturechangesin IPDepidemiologyshouldbeaccountedforwhendesigningvaccine recommendations.

TheaimofthisworkwastoestimatethenumberofIPDcases causedbyvaccineserotypesandNVTamongthe65+inthenear future(2014–2019)usinginterruptedtimeseriesanalysesbasedon population-widesurveillancedatafrom2004throughmid-2014.

Furthermore,weaimedtodeterminethenumberofpeopleneeded tobevaccinated(NNV;Kellyetal.,2004)withPCV13and/orPPV23 topreventoneIPDcaseperseasonandtocalculatethepotential publichealthimpactonIPDofscenarioswithdifferentlevelsof vaccineuptake.Althoughthequestionofwhichvaccinetousein olderadultshasbeenaddressedbyothers(Jiangetal.,2012;Smith etal.,2012;Vila-Corcolesand Ochoa-Gondar,2013;Fedsonand Guppy,2013;Jiangetal.,2014),toourknowledgethisstudyisthe ﬁrsttocomparedifferentvaccinestrategiesbypredictingthepublic healthimpactbasedondataobtainedinthePCV13era.

2. Methods 2.1. Datasources

InNorway,notificationofIPDismandatoryformicrobiological laboratoriesandmedicaldoctorsanditisassumedtohaveahigh and stablecoverage. We used theserotype-specificnotification datafromtheNorwegianSurveillanceSystemforCommunicable Diseases(MSIS;NorwegianInstituteofPublicHealth,2011)forIPD caseswithatestingdatebetween1January2004and30June2014 andaged65+.Datawereextractedon23July2014.AllIPDcases, definedasacaseinwhichS.pneumoniaewasisolatedfromanor- mallysterilesite,arenotifiabletoMSIS.Over98%areisolatedfrom bloodand/orCSFandmorethan90%ofisolatesareserotypedusing theQuellungreactionwithserotype-specificantisera(Vestrheim etal.,2010).AccordingtotheMSISregulations,theNIPHdoesnot requireethicalapprovalfortheuseofnotifieddataforthistypeof study.

StatisticsNorwayprovideddataonthenumberofNorwegian inhabitants per agegroup at the 1st of January of each corre- sponding year (Statistics Norway, 2014a), as well as predicted populationsizesforthefuture(StatisticsNorway,2014b).Weused thepredictedpopulationsizeatmediangrowthandusedlinear interpolationtodeterminemonthlypopulationsizesasdenomina- tor.In2014,Norwayhad5.1millioninhabitants,ofwhich821,558 (16%)were65+.

2.2. Interruptedtimeseriesanalyses

Wecategorisedourdatabyvaccine-type usingthefollowing designation:PCV13serotypes,serotypesthatarecoveredbyPPV23

butnotbyPCV13(PPV23–12),andNVT,deﬁnedasallserotypesnot coveredbyPCV13orPPV23;seeTable1.

Datawereaggregatedbymonths.Overall7%(305/4365)ofiso- latesmissed serotypeinformation;58% ofmissingsoccurredin 2004/2005.Imputationofmissingvalueswasperformedaccording tothedistributionofknownserotypesintherespectivemonth,the precedingandthefollowingmonth.ThenotiﬁedmonthlyIPDcases YtwereregressedusingaPoissonsegmentedtimeseriesanalysis incorporatingthechangesintheNorwegianchildhoodimmunisa- tionprogramme,thepopulationsizeandcorrectingforseasonality:

Yt=exp(log(Pop_t)+ˇ0+ˇ1×month_t+ˇ2×month^PCV7_t

+ˇ3×month^PCV13_t +ˇseasonal_k+εt) (1) wheremonth_tisthenumberofmonthsfromthestartofthestudy periodin January2004,and month^PCV7_t and month^PCV13_t arethe numbersofmonthsaftertheintroductionofPCV7vaccinationin July2006andtheswitchtoPCV13vaccinationinApril2011,respectively;beforetheinterventionsthesevariablesaresettozero.The termˇ0istheinterceptcoefﬁcient,ˇ1istheinitialtrend,ˇ2isthe changeintrendpostintroductionofPCV7vaccination,ˇ3 isthe changeintrendpostintroductionofPCV13vaccination,ˇseasonalk

istheseasonalityfactorvariableatmonthk=1,2..12,andεtisthe error.Toaccountforchangesinthepopulationsize,theNorwe- gianpopulationatmontht(Pop_t)wasusedasanoffset.Toaccount for additionalvariation, weincludeda dispersionparameter,, resultinginaquasi-Poissonmodel.

Themodel Eq. (1) wasfittedseparately toPCV13serotypes, PPV23–12serotypesandNVTfrom2004throughJune2014(126 months; Fig.1).Non-significant parameters (p≥0.05) weredis- cardedfromthefinalmodels;seeTable1.Datawereanalysedusing GLMwiththeMASSpackageinthestatisticalsoftwareRversion 3.1.0(SwissFederalInstituteofTechnologyZurich,2014).Thefit- tedmodelswereusedtopredictIPDcasecountsfortheperiodJuly 2014–June2019(60months).

SeasonsweredeﬁnedtorunfromJulytoJunethefollowing year(e.g.July2004–June2005).Seasonalcountswerecalculated by summing therespective monthlycounts. The PPV23counts werecalculatedbyaddingthepredictedvaluesofthePCV13and PPV23–12models.NotethatourPPV23countsthereforeinclude thecountsforPCV13serotype6A,whichisnotincludedinPPV23.

Duetolownumbers,itwasnotpossibletomodelthisserotype separatelyandhadlimitedeffectontheﬁnalmodel.Thefrequency of serotype6Adecreased froman averageof24 casesper sea- sonin2004–2009to11in2009/10andfurthertotwo6Acases in2013/14.

2.3. Thepredictionintervals

Wedeterminedthe95%predictionintervals(95%PI)bytak- ingintoaccountboththevariationrelatedtotheuncertaintyin theparameters(systematicpart)andtheuncertaintyrelatedtothe futuretrend.First,weperformedresidualbootstrapping(N=1000) andreﬁttedthemodelstoaccountforthesystematicpartofthe variation.Then,foreachofthe1000samples,weintroduceduncer- taintyinthepredictedtrendbyaddinga randomcomponentin termsofarandomwalkprocedure(Pearson,1905).Ineachtime step(month),arandomnormallydistributednumberwithmean zeroandstandarddeviation0.00137wasdrawnandthecumulated valueswereexponentiatedandaddedtothepredictedvalues.This procedurewasrepeated1000timesforeachbootstrapsampleand the95%PIwasobtainedasthe2.5%and97.5%valuesofthesorted countsineachseason.Thevalueof0.00137waschosenbasedon preliminarysimulationsandcorrespondedtoachangeof5%inthe trendofthePCV13counts.

(3)

Table1

Overviewofthevariablesandtheirparametersofeachvaccine-typespeciﬁcquasi-Poissonregressionmodel.Ifnoparameterestimateispresented,thevariablewasdiscarded fromtheﬁnalmodel.Themonthlydenominatorwasincludedinallmodelsasoffset.

PCV13 PPV23-12 NVT

Includedserotypes 1,3,4,5,6A,6B,7F,9V,

14,18C,19A,19F,23F

2,8,9N,10A,11A,12F, 15B/C,17F,20,22F,33F

Allserotypesnotincludedin PCV13orPPV23

Variables Estimate pvalue Estimate pvalue Estimate pvalue

Interceptˇ0 −9.679 <2exp-16 −11.447 9.06exp-06 −12.153 <2exp-16

TimetrendfromJanuary2004ˇ1 0.006 <2exp-16

Changeintime-trendfromPCV7introductionˇ2 −0.015 <2exp-16 0.015 1.89exp-08

Changeintime-trendfromtheswitchtoPCV13ˇ3 −0.012 0.009 −0.011 0.011 −0.015 1.59exp-02

ˇsesonalkJanuary Reference Reference Reference

February −0.222 0.023 −0.229 0.092 −0.384 0.061

March −0.071 0.445 −0.197 0.142 −0.019 0.917

April −0.089 0.345 −0.165 0.213 −0.323 0.106

May −0.174 0.072 −0.269 4.96exp-02 −0.093 0.618

June −0.483 1.18exp-05 −0.561 2.37exp-04 −0.441 3.33exp-02

July −0.814 4.87exp-10 −0.689 3.60exp-05 −0.710 2.94exp-03

August −1.222 2.52exp-14 −0.978 0.000 −0.814 1.02exp-03

September −0.260 0.011 −0.605 0.000 −0.408 0.055

October −0.120 0.217 −0.479 0.002 −0.381 0.069

November −0.099 0.309 −0.316 0.027 −0.436 0.041

December 0.119 0.196 0.115 0.360 0.106 0.560

Dispersionparameter 1.250 0.971 1.203

Residualdeviance(degreesoffreedom) 141.25(112) 0.032 117.32(112) 0.347 138.67(112) 0.044

2.4. Numberneededtovaccinate

TheNNVisanalogoustothenumberneededtotreat(Laupacis etal.,1988).Inthepresentcontext,theNNVindicatesthenumber ofpeopleaged65+whoneedtobevaccinatedina certainsea- sontopreventoneIPDcaseinthatsameseasonandagegroup.

WedeterminedtheNNVforthreevaccinationstrategies,PPV23- only,PCV13-onlyoruseofPCV13andPPV23combined,usingthe methoddescribedbyKellyetal.,2004:

NNV=1/Inc_unvac×VE (2)

whereIncunvacistheseasonalincidenceintheunvaccinatedpop- ulation,whichwascalculatedbydividingthepredictedseasonal vaccine-typecount, ˆYs,bythesizeoftheunvaccinatedpopulation;

seeEq.(3).Thesizeoftheunvaccinatedpopulationwasdetermined bycorrectingthepredictedpopulationsize,Pops,fortheassumed immunisedpopulation:

Incunvac=Yˆs/(Pop_s−(Pop_s×PPV23 uptake×VE PPV23)) (3) ThePPV23uptakewasassumedat22%(medianoftheestimated vaccineuptakeinNorway;unpublisheddataNIPH).WeusedVE 60%(range40–70)for PPV23(Table2), and 75%(range55–90) forPCV13guided byresultsof theCAPiTAstudy(Bontenetal., 2014).Inthecombinedscenario(PCV13+PPV23),weusedtheVEof PCV13forthePCV13serotypesandthePPV23-VEforthePPV23–12 serotypes.We usedthepointestimatesforVEtodeterminethe mostplausibleNNVandthelowerand upperbounds todeter- mineitsuncertainty.ItisknownthattheVEislowerforsomerisk groupscomparedtothe65+populationwithoutcomorbidities(see forreviewSteensetal.,2014a).However,sincethisstudyfocused onthegeneral65+populationthiswasnottakenintoaccount.

2.5. Publichealthimpactofincreaseduptakeofthedifferent vaccines

We determined the publichealth impact per future season (PHIs)bycalculatingtheadditionalnumberofcasespreventedby thethreevaccinationstrategies,PPV23,PCV13orPCV13+PPV23, atdifferentlevelsofvaccineuptake,whichwouldnothavebeen preventedifthecurrentscenariowithPPV23-onlyusedby22%of the65+wouldcontinue.

PHI_s=(Scenario level uptake−initial uptake)×VE×Yˆs (4) ThePHIs forthecombinedscenariowasdeterminedbycom- biningthePHIsforPCV13withthePHIsforPPV23–12.Thescenario levelsforvaccineuptakewere28%,36%or45%inthe65+population.Thesescenarioswerepartlybasedontheestimatedinﬂuenza vaccineuptake in Norwayin 2012/2013(28%without underly- ingcondition,36%among thegeneral65+ population;personal communicationKMRydland,NIPH,11April2014).Similartothe calculationofNNV,weusedthepointestimatesofthepredicted seasonalcountsandVEtocalculatethemostplausiblePHI_s,while thelowerandupperboundswereusedtoquantifyitsuncertainty.

Wedidnottakeintoaccountwaningimmunity,butinsteadwe usedaconstantvalueofVEforallﬁvepredictedseasons.Thedura- tionofprotectionofPPV23isconsideredtobelongerthan5years (Andrewsetal.,2012;Ochoa-Gondaretal.,2014),whilecurrently nodataisavailableforPCV13.WeassumedthatPCV13vaccina- tionwouldbeinitiatedinthe65+in2014/15,andthatthelevel ofuptake ofthevaccines wouldbereachedwithinthat season, withtheexceptionforPCV13becauseofthefollowing:inNorway, PCV13isrecommendedtobegivenatleast3yearsafterthelast PPV23;PPV23shouldberepeatedevery10years (Steensetal., 2014a).Peoplewhowerevaccinatedwithinthelastthreeseasons

Fig.1.Graphicalrepresentationoftheperiodwithobservedandwithpredicteddata.ThesyringespresentthetimethatPCV7wasintroduced(July2006)andtheswitchto PCV13(April2011).

(4)

Table2

LiteratureoverviewofestimatesoftheeffectivenessofPPV23topreventvaccine-typeIPDinthe65+.Ifdifferentclinicalpicturesoragegroupswereusedinthestudy,this isindicatedinthetablebetweenbrackets.Notethatmostoftheempiricalstudiesthatarepresentedarealsoincludedinthemeta-analyses..

Reference Kindofstudy Studyperiod VE(95%C.I.)

Butleretal.,1993 Indirectcohortanalysison surveillancedata

1978–1992 70(30–78)[immune-competent65–74years]

78(54–89)[immune-competent≥75years]

Shapiroetal.,1991 Case-controlstudy 1984–1990 80(51–92)/71(30–88)[65–74years;<3years/3–5yearsafter vaccination]

67(20–87)/53(−15to88)[75–84years;<3/3–5years]

46(−31to78)/22(−90–68)[≥85years;<3/3–5years]

Honkanenetal.1999 Non-randomisedtrial 1993–1994 60(−40to90)

Jacksonetal.,2003 Retrospectivecohortstudy 1998–2001 44(7–67)[bacteraemia]

Andrewsetal.2004 Indirectcohortanalysison surveillancedata

1997–1998, 2001–2002

79(−14to96)

Dominguezetal.2005 Case-controlstudywith hospitalcontrols

2001–2002 64(31–82) Mooneyetal.2008 Screeningmethodon

surveillancedata

2003–2004 62(45–73)[all-serotypeIPD]

Vila-Corcolesetal.2010 Case-controlstudy 2002–2007 77(40–92)[population60+]

Andrewsetal.,2012 Indirectcohortanalysison surveillancedata

1998–2010 48(32–60)[<2yearsaftervaccination]

24(10–36)[overall]

Hutchisonetal.1999 Meta-analysis 1966–1996 83(69–91)

MelegaroandEdmunds,2004 Meta-analysis 1985–1999 65(−49to92)

Puig-Barberaetal.2002 Meta-analysis 1964–2000 32(−18to61)

Mooreetal.2000 Meta-analysis 1966–2000 47(−94to86)[bacteraemia]

werethereforenotyeteligibletoreceivePCV13.Theresultinglev- elsofvaccineuptakethathavebeenusedinEq.(4)arepresented inTable3.

AnalysesweredoneinR(http://www.r-project.org/),Stata13 andExcel2010.

3. Results

3.1. ChangeinIPDtrendsfollowingchildhoodvaccination

Duringthestudyperiod,4365IPDcasesaged65+werenoti- fiedinNorway,correspondingto49%(4365/8847)ofallnotified IPDcases.TheIPDcasecountsshowedclearseasonalitywithhigh countsinthewinterandsignificantlylowercountsinthesummer (Fig.2;Table1).

Theaveragenumberof65+casesperseasondecreasedfrom 498(incidence73/100,000)in2004–2006beforetheintroduction ofPCV7to400(54/100,000)in2010/11beforeswitchingtoPCV13.

Thisdroppedfurtherto274(34/100,000)in2013/14;seeTable4.

ThepercentagesofIPDcasescausedbyPCV13andPPV23serotypes decreasedfrom80%and89%in2004/05,respectively,to26%and 64%in2013/14.

Overall,themodelsﬁttedwelltheobservedmonthlyIPDcase counts (Fig. 2). The estimated number of IPD cases caused by PCV13serotypesdecreasedsigniﬁcantlyfromthemomentPCV7 wasintroduced,andtheswitchtoPCV13ledtoafurtherdeclinein cases(Fig.2;Table1).BothPPV23–12andNVTIPDcountsincreased

slightlyintheperiodbeforetheswitchtoPCV13.Afterswitching toPCV13,theincreasewashaltedforNVT-IPD,andthePPV23–12 IPDcountsdecreasedslightly.

3.2. Trendsforthenearfuture

OurmodelspredictedacontinueddecreaseinPCV13-IPDcounts inthenearfuture,withanestimateof17(95%PI6–52)casesin 2018/19, representing8%of thetotal IPDcasecounts(Table4).

At that time, thePPV23–12 IPDcount is predicted at 98 (95%

PI 33–295)cases. Taken together, the resultssuggest a declin- ingtrendinPPV23-IPDcountsto52%ofthetotalcasecountsin 2018/19;however,thepredictionintervalsarewideandinconclu- sive(Table4).NVT-IPDcountsmayshowasmallincreaseto104 cases(95%PI35–313)in2018/19,thenaccountingfor48%ofthe cases.

3.3. NumberneededtobevaccinatedtopreventoneIPDcase Despitea higherVEforPCV13thanforPPV23,theNNVwas largerforPCV13becauseofitslowervaccine-typeIPDincidence (Table5).To preventone IPDcase in 2014/15, 16,524 persons (uncertainty 13,770–22,533)shouldbe vaccinated withPCV13- only or 7149 (6128–10,724) with PPV23-only. This value will increaseto60,910(50,759–83,060)and11,508(9864–17,262)in 2018/19,respectively;a5.3timesdifferencebetweenPCV13and PPV23.ThecombinationofPCV13andPPV23willhavethelowest

Table3

Exampleofthecalculationofthelevelofvaccineuptake(=scenario–initialuptake)thatwasusedtopredictthepublichealthimpactofPPV23,PCV13orPPV23+PCV13.

Hereweonlypresentthescenariolevelof36%.NotethatPCV13isrecommendedtobegivenatleast3yearsafterthelastPPV23;PPV23shouldberepeatedevery10years (Steensetal.,2014a).

PPV23 PCV13 PCV13+PPV23

2014/15 14%PPV23^a 29.4%PCV13^b 29.4%PCV13and14%PPV23–12

2017/18 14%PPV23^a 36%PCV13 36%PCV13and14%PPV23–12

2018/19 14%PPV23^a 36%PCV13 36%PCV13and14%PPV23–12

a14%=scenariolevel36%–initialuptakeof22%.

bThePCV13level=scenariolevel36%–thepopulationthatisnotyeteligibleforPCV13vaccinationbecauseofrecentPPV23vaccination.Intheﬁrstseason,3/10ofthe populationthathasinitiallybeenvaccinated(22%)isineligibleforvaccination,thesecondseason2/10×22%andthethirdseason1/10×22%.

(5)

Fig.2.Theobserveddata(ingreen)andpredictedmonthlynumberofIPDcases(inblue)with95%predictionintervals(inred).A:PCV13,B:PPV23–12,C:PPV23(including serotype6A),D:NVT.Thearrowsindicatethetimingofvaccineintroduction.

NNV,indicatingthelargestpreventiveeffect.Becauseofthefurther decreaseinvaccine-typeIPDincidence,theNNVwillincreaseover timeforallscenarios.

3.4. Publichealthimpactofincreaseduptakeofthedifferent vaccines

ThecombinationofPCV13andPPV23isfoundtopreventmost casesduringallpredictedseasonsandatallinvestigatedlevelsof vaccineuptake(Fig.3),althoughthedifferencecomparedtoPPV23- onlyispredictedtobeonlythreecasesin2018/19.Increasingthe vaccineuptakewillleadtoalargerpublichealthimpact;anincrease

inPPV23uptakeaffectsthepublichealthimpactmoresubstantially thananincreaseinPCV13uptake,mainlyasaresultsofthesteeper decreaseinPCV13-IPDincidencecomparedtoPPV23–12IPDinci- dence.Theimpactofallscenarioswilldecreaseovertimeasaresult ofthedecreaseinvaccine-typeIPDcounts.

Thepublichealthimpactofascenariowherevaccinationwith PCV13isadded tothe22%of65+ thatarecurrentlyvaccinated withPPV23isestimatedtopreventadditionally7(5–8)casesin 2014/15,which is similartotheexpectedimpactfromincreas- ingthePPV23-onlyuptaketo28%,whichmaypreventfurther6 (4–7)cases.Thesenumberswillhavedecreasedto3(2–3)and4 cases(3–5)in2018/19,respectively.Atahigherlevelofvaccine

Table4

ObservedandpredictednumberofIPDcasesaged≥65yearperseason(July–June),byvaccine-type.BehindtheobservednumberofIPDcasesinthesurveillancedata,the percentageofallIPDcasesinitsseasoncausedbyserotypescoveredbytherespectivevaccinesispresentedinbrackets.Forthepredictedresults,the95%predictioninterval (95%PI)ispresentedinbrackets.

Surveillancedata PCV13(%) PPV23–12(%) PPV23(%) NVT(%)

Pre-vaccination 2004/05 416(80) 61(12) 461(89) 43(8)

2005/06 377(79) 74(16) 431(91) 25(5)

PCV7period 2006/07 395(82) 65(13) 423(88) 23(5)

2007/08 292(66) 101(23) 370(84) 49(11)

2008/09 271(64) 106(25) 347(81) 49(12)

2009/10 179(50) 109(30) 277(77) 69(19)

2010/11 205(51) 108(27) 303(76) 88(22)

PCV13period 2011/12 162(45) 117(32) 274(76) 83(23)

2012/13 118(36) 133(40) 247(74) 82(25)

2013/14 71(26) 106(39) 175(64) 97(35)

Predictedresults PCV13(95%PI) PPV23–12(95%PI) PPV23^a(95%PI) NVT(95%PI)

PCV13period 2014/15 58(55–61) 110(105–116) 168(162–175) 95(90–101)

2015/16 43(35–54) 107(86–133) 150(127–178) 98(78–122)

2016/17 32(20–50) 104(66–164) 135(96–199) 100(63–159)

2017/18 23(11–50) 100(48–214) 124(70–241) 102(48–218)

2018/19 17(6–52) 98(33–292) 115(49–313) 104(35–313)

aNotethatforthepredictedresults,PPV23includesserotype6a,asthisestimatewasdeterminedbasedonthesumofthepredictionbymodelsPCV13andPPV23–12.

(6)

Table5

NumberneededtovaccinatetopreventoneIPDcaseperseason(July–June),presentedbyvaccine-type.AsVEweusedforPPV23:60%(range40–70),andforPCV13:75%

(range55–90).Theuncertainty,presentedbetweenbrackets,isdeterminedbytheuncertaintyinVE.

Season PPV23^a PCV13 PCV13+PPV23

2014/15 7149(6128–10,724) 16,524(13,770–22,533) 6580(5577–9492)

2015/16 8223(7048–12,334) 22,896(19,080–31,222) 7672(6514–11,135)

2016/17 9315(7,985–13,973) 31,725(26,438–43,261) 8799(7482–12,842)

2017/18 10,413(8926–15,620) 43,959(36,632–59,944) 9942(8467–14,583)

2018/19 11,508(9864–17,262) 60,910(50,759–83,060) 11,089(9456–16,336)

aIncludingserotype6A.

Fig.3.Thepublichealthimpact:thepredictednumberofIPDcasespreventedbyPPV23(darkgrey),PCV13(lightgrey)andacombinationofPCV13andPPV23(grey)in comparisontothecurrentscenario(PPV23-onlyat22%uptake).A:28%vaccineuptake,B:36%uptake,C:45%uptake.Theerrorbarsindicatetheuncertaintyinthepublic healthimpactbasedontheuncertaintyinVE.

uptake,anincreaseinPPV23-onlyuptakewillhavealargereffect thanaddingPCV13,andPPV23willaccountforalargerproportion oftheimpactofthecombinedregimen(Fig.3).

4. Discussion

WehaveshownthattheimplementationofPCVinthechild- hoodimmunisation programmehassubstantially decreasedthe PCV13-IPDincidenceamongthe65+ asaresultofindirectpro- tection.ThenumberofIPDcaseswilllikelycontinuetodecrease further,withlimitedornoincreaseinNVT-IPD.TheNNVforPCV13 willconsequentlyincreaseconsiderably.ThepotentialforPCV13 topreventIPDinthe65+inNorwaywillthereforebelimited.In thelongerrun,increasingPPV23uptakewillhaveahigherpublic healthimpactthanaddingPCV13toPPV23withunchangedlev- elsofvaccineuptake.AnincreasedvaccineuptakewithPCV13and PPV23combinedwillpreventmostcases,althoughtheadditional effectofPCV13willdecreaseandisonlymarginalin2018/19.

ThisstudyfocusedonIPD,whilethehighestburdenofpneu- mococcalinfectionsisonlowerrespiratorytractinfections/non- invasivepneumonia.Canweextrapolatetheresultsfromthisstudy tonon-invasivepneumonia?AlthoughPPV23isthoughttobeeffec- tiveinpreventingIPDinolderindividuals,particularlyinhealthy individualsunder75years(Moberleyetal.,2013;Andrewsetal., 2012),itsprotectionagainstnon-invasivepneumoniaisdebated (Moberleyetal.,2013;MelegaroandEdmunds,2004;Jacksonetal., 2003;Ochoa-Gondaretal.,2014).PCV13canpreventabouthalf ofthePCV13-pneumoniacasesaged65+(Bontenetal.,2014).So, theremightbea potentialfor PCV13inpneumoniaprevention.

ThereisnoinformationavailableabouttheincidenceofPCV13- pneumoniainNorway.Resultsofalargetrialamong65+inthe Netherlandsestimatedthatabout14%ofallcommunityacquired pneumoniacaseswerecausedbyPCV13serotypes(Bontenetal., 2014).NotethatintheNetherlands,the10-valentPCVinsteadof PCV13isusedinchildren.PCV13-serotypeshavebeenshowntobe lessoftenthecauseofnon-invasivecomparedtoinvasivepneumo- coccalpneumonia(Benﬁeldetal.,2013).Togetherwiththefurther decreaseinPCV13-IPDincidence,thissuggestsalsoalimitedrole forPCV13inpneumoniapreventioninNorway.Still,moreresearch

isneededontheincidenceofPCV13-pneumoniatoestimateitsreal preventivepotentialinthe65+.

Weusedasimplestatisticalpredictionmodelfittedonnotifica- tiondata.Weexploredseveralwaysofimplementingthechanges inthechildhoodimmunisationprogrammeinthemodel,including constantinterventioneffectsandagradualbuild-upoftheinter- ventioneffect.Asthemorecomplexmodelsdidnotsignificantly improvethemodelfit,wedecidedonusingasimpleapproachasa timetrendstartingatthemomentofintervention.Wemodelledthe vaccine-typeserotypesingroupsinsteadofindividualserotypes asthishasbeenshowntoprovidethebestestimatesforchanges inIPDincidence(Weinbergeretal.,2013).Theadvantageofour approach is thatwe didnotneed tomake assumptions onthe sizeoftheindirecteffectandserotypereplacement,asthesewere intrinsictothedata.Still,thepredictionsarebasedonextrapo- lationofcurrenttrends.Topreventtoostrictconclusionsforthe future,weaddedstochasticitytoourpredictionintervalsthrough arandomwalkprocess.Dataonpneumococcalcarriageinchildren, themaintransmittersofpneumococci,indicatethatthesubstan- tialindirectprotectionofPCV13isindeedlikelytocontinueinthe future(Leeetal.,2014;Gounderetal.,2014;vanHoeketal.,2014;

Ricketsonetal.,2014;Steensetal.,2014b).Itisunlikelythatuse ofPCV13amongthe65+willinduceindirectprotectionbecause ofthelowcarriageratesamongolderagegroups(<5%;Flamaing etal.,2010).Theverylimitedorabsenceofserotypereplacement aspredictedbyourmodelsislesscertain,asreﬂectedinthewide 95%PIs.Severalstudiesindicatethatserotypereplacementmaybe occurringafterswitchingtoPCV13,thoughtoalesserdegreethan thatobservedafterPCV7introduction(Mooreetal.,2014;Steens et al., 2013;Kaplan et al., 2013). As it was shown after PCV7 implementation that serotype replacement was complete only afterafewyears(Hanageetal.,2010),serotypereplacementmight stilloccur.Ifserotypereplacementwillstilloccur,dependingonthe replacingserotypes,PPV23mayhavelargerpublichealthimpact thanpredictedbythisstudy.

The NNV and public health impact are dependent on the assumption on the initial level of vaccine uptake. The vaccine recommendations for risk-groups changed in May 2013, such thatpatientswithcertaincomorbiditiesarerecommendedtouse PCV13inadditiontoPPV23,butforthegeneral65+population,

(7)

PPV23isstilltheonlyrecommendedvaccine(Steensetal.,2014a).

Althoughtheexactvaccineuptake among the65+ isunknown, data on the number of vaccines sold for common risk-groups (includingthe65+)indicatethatPPV23uptakehasnotchanged substantially (unpublished data NIPH). If we have underesti- matedthecurrentvaccineuptakeforPPV23,theestimatedNNV wouldbehigher forPPV23but lowerfor both, thePCV13-only scenario and the scenario with combined use of PCV13 and PPV23. Furthermore, the public health impact of PPV23-only andthecombinedscenariowouldbelower,whiletheimpactof PCV13wouldbasicallybeunchanged.Notethatallpublichealth impact scenarios are presented in comparison to the baseline scenariowherePPV23wasusedatalevelof22%.Unfortunately we were not able to compare thepublic health impact of the different vaccinationscenarios to a scenario in the absence of anypneumococcalvaccination,asweestimatedthefutureinci- dencebyﬁttingnotiﬁcationdatafromaperiodwherePPV23had beenrecommended for the65+. The publichealth impact of a PCV13-onlystrategywouldbeslightlylargerinascenariowhere PPV23vaccination were to cease. However, as theduration of protectionfromPPV23amongthosealreadyvaccinatedisassumed tolastlongerthanthenumberofyearswepredictedforinthis study,theunderestimationfromusingaPPV23baselineislikely limited.

Severalcountrieshavelikewiseinvestigatedthebestapproach toprevent IPDin older adults. Population-based Markov mod- ellingstudiesfromGermanyandUKconcludedthatvaccination withPPV23remainstheoptimalvaccinationstrategyfrompub- lichealthandbudgetperspectives(Jiangetal.,2012;Jiangetal., 2014).AMarkovstate-transitionmodelusedonUSdatapredicted thatPCV13willhavemoreimpactthanPPV23,butonlyincaseof moderateindirectprotectionbychildhoodimmunisation(Smith etal.,2012).AstaticcohortmodelonUSdatashowedthatacom- binedstrategywouldpreventmostcases(Chenetal.,2014).Astudy usingamoresimilarapproachtoours(Poissonmodel;Link-Gelles etal.,2013)concludedonastablerateofPCV13-IPD,suggesting thatPCV13mighthaveastablepreventivepotentialinthe65+up to2020.Noneofthesemodelsimplementedsurveillancedataofthe PCV13era,andallweresensitivetothesizeoftheindirecteffectof childhoodimmunisation.Asweusedsurveillancedatauptothree yearsaftertheswitchtoPCV13,wecouldimplementtherealsize oftheindirecteffectuptomid-2014,andshowedthatthesubstan- tialindirectprotectionlikelydecreasesthepublichealthimpactof PCV13inNorway.

Pneumococcican be an importantcause of super-infections during influenza infections (O’Brien et al., 2000;Kuster et al., 2011;Fleming-Dutraetal.,2013;Talbotetal.,2005).Wetherefore evaluatedtheeffectofinfluenzaillness(includingthepandemic season)onourestimatesbyaddingtheincidenceofinfluence-like illness (ILI) determined by the ILI sentinel surveillance (Hauge etal.,2009)tothemodels,butthisdidnotsignificantlyimprove themodelsifwealreadycorrectedforseasonality.Still,influenza vaccinationhasbeenshown toplaya rolein thepreventionof IPD (Christenson et al., 2004;Gilchrist et al., 2012), and in an optimalstrategy for IPD prevention, pneumococcalvaccination is combined with influenza vaccination. Increasing the uptake of both vaccines is needed. Healthcare providers should be involvedinencouraginguptakeofvaccination(Schneebergetal., 2014).

5. Conclusions

Althoughourresultsshouldbeinterpretedwithcautiondueto theirpredictivenature,thisstudysuggeststhatinNorwayvaccine- typeIPDinthe65+willcontinuetodecrease.Furtherreductions

in IPD in the era of childhood vaccination can beachieved by increasingthePPV23uptake.Still,astheNNVwaslowestforthe combinationof PCV13andPPV23,theuseof PCV13andPPV23 combinedshouldbeconsideredforthemostfragile population, includingthe65+withcomorbidities,asiscurrentlyrecommended inNorway.Asthepreventivepotentialofthecurrentlyavailable pneumococcalvaccinesispredictedtodecrease,vaccinesconfer- ringbroaderprotectionshouldbedeveloped.

ConﬂictofInterest Nonedeclared

Financialsupport

NorwegianInstituteofPublicHealth

Authorcontributions

AS,DFVandBFdBcontributedtotheconceptionofthestudy andthestudydesign,BFdBwrotetheinitialR-syntax,ASandBFdB performedtheﬁnaldataanalysis,AS,DFVand BFdBinterpreted theresults,ASdraftedthemanuscript,andBFdBandDFVcritically revisedthemanuscript.Allauthorshaveseenandapprovedthe ﬁnalmanuscript.

Acknowledgments

Weacknowledgeallmedicalmicrobiologicallaboratoriesand cliniciansinNorwayfornotiﬁcationtoMSISandsendingofisolates totheNationalReferenceLaboratoryattheNIPH.WethankAnne RamstadAlme,GunnhildRødalandLeneHaakensen(NIPH)forall laboratoryanalysesandKirstenKonsmo(NIPH)fordataentryofall IPDcasesinMSIS.WeacknowledgeAliciaBarrasaoftheEuropean CentreforDiseasePreventionandControlandKarinNygårdofthe NIPHfortheirfeedbackonthestudyprotocol.WearegratefultoJon MichaelGranoftheInstituteofBasicMedicalSciences,University ofOsloforhisadviceonthecalculationofthe95%PI.

References

Andrews,R.M.,etal.,2004.Effectivenessofapubliclyfundedpneumococcalvac- cinationprogramagainstinvasivepneumococcaldiseaseamongtheelderlyin Victoria.Australia.Vaccine23,132–138.

Andrews,N.,etal.,2012.Impactandeffectivenessof23-valentpneumococcal polysaccharidevaccineagainstinvasivepneumococcaldiseaseintheelderly inEnglandandWales.Vaccine30,6802–6808.

Benﬁeld,T.,etal.,2013.Serotypedistributioninnon-bacteremicpneumococcal pneumonia:associationwithdiseaseseverityandimplicationsforpneumococ- calconjugatevaccines.PLoSOne8,e72743.

Bonten,M.,Bolkenbaas,M.,Huijts,S.,Webber,C.,Gault,S.,Gruber,W.,andGrobbee, D.,12-3-2014.Communityacquiredpneumoniaimmunisationtrialinadults (CAPiTA).Pneumonia,volume3:ISPPDspecialissue.Page95.

Butler,J.C.,etal.,1993.Pneumococcalpolysaccharidevaccineefﬁcacy.Anevaluation ofcurrentrecommendations,JournaloftheAmericanMedicalAssociation270, 1826–1831.

Chen,J.,etal.,2014.Cost-effectivenessofpneumococcalvaccinesforadultsinthe UnitedStates.Adv.Ther.31,392–409.

Christenson,B.,etal.,2004.Additivepreventiveeffectofinﬂuenzaandpneumococ- calvaccinesinelderlypersons.Eur.Respir.J.23,363–368.

Dominguez,A.,etal.,2005.Effectivenessofpneumococcalvaccinationforelderly peopleinCatalonia,Spain:acase-controlstudy.ClinInfectDis40,1250–1257.

European Medicines Agency. 30-5-2013. Assessment report for Prevnar13 –updated.http://www.ema.europa.eu/docs/enGB/documentlibrary/EPAR-

AssessmentReport-Variation/human/001104/WC500152448.pdf.

Fedson,D.S.,Guppy,M.J.,2013.Pneumococcalvaccinationofolderadults:conjugate orpolysaccharide?Hum.Vaccin.Immunother.9,1382–1384.

Fleming-Dutra,K.E.,etal.,2013.Effectofthe2009InﬂuenzaA(H1N1)pandemicon invasivepneumococcalpneumonia.J.Infect.Dis.207,1135–1142.

Flamaing,J.,etal.,2010.Pneumococcalcolonizationinolderpersonsinanonout- breaksetting.J.Am.Geriatr.Soc.58,396–398.