Article
Reference
Baseline polysaccharide-specific antibodies may not consistently inhibit booster antibody responses in infants to a serogroup C
meningococcal protein-polysaccharide conjugate vaccine
BLANCHARD ROHNER, Géraldine, et al .
Abstract
Negative correlations between baseline antibody concentrations and increases in antibody concentrations (after booster doses of vaccines) have been reported previously. Such correlation coefficients are widely reported by statisticians to be subject to mathematical coupling. Negative correlations may be attributable partly or wholly to the combination of mathematical coupling and measurement error (or other short term fluctuations in measurements) and therefore not clinically interpretable. In this study we re-analysed the serum antibody responses from five clinical trials of serogroup C meningococcal conjugate vaccine (MenCV) given to infants for priming followed by boosting with MenCV or a meningococcal A/C polysaccharide vaccine (MenA/C) at 12 months of age. Using Pearson's correlation method to assess the effect of pre-booster MenC-IgG concentration on the relative increase in MenC-IgG concentration post-booster, a significant negative correlation was observed for all the groups, indicating that high pre-booster antibody was associated with a smaller rise in antibody post-booster. We tested two additional statistical [...]
BLANCHARD ROHNER, Géraldine, et al . Baseline polysaccharide-specific antibodies may not consistently inhibit booster antibody responses in infants to a serogroup C meningococcal protein-polysaccharide conjugate vaccine. Vaccine , 2012, vol. 30, no. 28, p. 4153-4159
PMID : 22554466
DOI : 10.1016/j.vaccine.2012.04.063
Available at:
http://archive-ouverte.unige.ch/unige:149071
Disclaimer: layout of this document may differ from the published version.
1 / 1
ContentslistsavailableatSciVerseScienceDirect
Vaccine
j o ur na l ho me p ag e : w w w . e l s e v i e r . c o m / l o c a t e / v a c c i n e
Baseline polysaccharide-specific antibodies may not consistently inhibit booster antibody responses in infants to a serogroup C meningococcal
protein–polysaccharide conjugate vaccine
Geraldine Blanchard-Rohner
a,b,∗, Hilary Watt
c, Dominic F. Kelly
a, Ly-Mee Yu
c, Matthew D. Snape
a, Andrew J. Pollard
aaOxfordVaccineGroupandtheNIHROxfordBiomedicalResearchCentre,UniversityofOxford,CentreforClinicalVaccinologyandTropicalMedicine,Oxford,UK
bDepartmentofPaediatrics,Children’sHospitalofGeneva,UniversityHospitalofGeneva,Switzerland
cCentreforStatisticsinMedicine,UniversityofOxford,Oxford,UK
a r t i c l e i n f o
Articlehistory:
Received25October2010
Receivedinrevisedform12April2012 Accepted19April2012
Available online 1 May 2012
Keywords:
MenC
Baselineantibody Responsetobooster
Protein–polysaccharideconjugatevaccine
a b s t r a c t
Negativecorrelationsbetweenbaselineantibodyconcentrationsandincreasesinantibodyconcentra- tions(afterboosterdosesofvaccines)havebeenreportedpreviously.Suchcorrelationcoefficientsare widelyreportedbystatisticianstobesubjecttomathematicalcoupling.Negativecorrelationsmaybe attributablepartlyorwhollytothecombinationofmathematicalcouplingandmeasurementerror(or othershorttermfluctuationsinmeasurements)andthereforenotclinicallyinterpretable.Inthisstudy were-analysedtheserumantibodyresponsesfromfiveclinicaltrialsofserogroupCmeningococcalcon- jugatevaccine(MenCV)giventoinfantsforprimingfollowedbyboostingwithMenCVorameningococcal A/Cpolysaccharidevaccine(MenA/C)at12monthsofage.UsingPearson’scorrelationmethodtoassess theeffectofpre-boosterMenC-IgGconcentrationontherelativeincreaseinMenC-IgGconcentration post-booster,asignificantnegativecorrelationwasobservedforallthegroups,indicatingthathighpre- boosterantibodywasassociatedwithasmallerriseinantibodypost-booster.Wetestedtwoadditional statisticalmethodsthataccountformathematicalcoupling.UsingBlomqvistmethodofadjustmentto assesstheplausibleextentofbias,correlationcoefficientswerestillnegativeprovidingerrorvariance waslow.Theothermethod,amultilevelmodellingspecificationofOldham’smethodappearednottobe appropriate.Incontrast,usingPearson’scorrelationmethodaconsistentnegativecorrelationbetween carrierprotein-specificbaselineantibodyconcentrationandtheincreaseinMenC-specificantibodycon- centrationwasonlyobservedfollowingboosterimmunisationwiththeprotein-polysaccharideconjugate vaccinebutnotfollowingtheMenA/Cplainpolysaccharidevaccine.Thesefindingssuggestthatanalysis oftheinhibitoryeffectofbaselineantibodyontheresponsetoboosterimmunisationischallengingand shouldaccountforthepossibilityofmathematicalcouplingandmeasurementerror.Thataninhibitory effectofbaselineantibodycannotbeassumedaprioriissupportedbyobservationsinanimalmodels, whichshowthatbaselineantibodycanbothsuppressorenhancetheantibodyresponsetoaspecific antigen.
© 2012 Elsevier Ltd. All rights reserved.
Abbreviations: CRM197,amutantofdiphtheriatoxoidusedascarrierprotein indiverseproteinpolysaccharideconjugatevaccines;GMC,geometricmeancon- centration;Hib,Haemophilusinfluenzaetypeb;MenC,meningococcalserogroupC polysaccharide;MenA/C,meningococcalAandCpolysaccharidevaccine;MenCV, serogroup C meningococcal conjugate vaccine; MMR, measles, mumps and rubellavaccine;9vPnC-MenCV,combined9-valentpneumococcal/meningococcal serogroupCconjugatevaccine;DT,diphtheriatoxoid;SD,standarddeviation;TT, tetanustoxoid.
∗Correspondingauthorat:CentreforClinicalVaccinologyandTropicalMedicine, ChurchillHospital,Oxford,OX37LJ,UK.Tel.:+441865857444;
fax:+441865857420.
E-mailaddress:geraldine.blanchardrohner@paediatrics.ox.ac.uk (G.Blanchard-Rohner).
1. Introduction
Vaccinesarefrequentlyadministeredtoindividualswithpre- existingserumantibodyspecificforthevaccineantigen.Thisoccurs becauseofeither(i)presenceoftransplacentallyacquiredmater- nal antibodyin infants, (ii)the existence of naturallyacquired antibodyfollowingexposuretotheorganismfromwhichthevac- cineisderived (orrelatedorganisms)throughcarriage,(iii) the administration of prior doses of vaccine, or (iv) in the case of polysaccharideconjugatevaccines,thepriorand/orconcomitant administrationofvaccinescontainingtheproteincarrier(i.e.rou- tine diphtheriatoxoid,tetanustoxoidandpertussisvaccinesin, forexample,Hib-TT,MenC-TT/-CRM197andPnC-TT/-DT/-CRM197 0264-410X/$–seefrontmatter© 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.vaccine.2012.04.063
4154 G.Blanchard-Rohneretal./Vaccine30 (2012) 4153–4159
antigens)thatisalsopresentinthevaccine.Variousstudiesinani- malshavereportedthatthepresenceofantibodyatbaselinecan eithersuppressorenhancetheantibodyresponsestothespecific antigentowhichtheyaretargeted[1–3].Possiblemechanismsfor enhancementofantibodyresponsebybaselineantibodyinclude (1)thecaptureofsmallquantitiesofantigenbypre-existingIgG (especiallysubclassesIgG1,IgG2aandIgG2b)whichthentargets theantigentoantigen-presenting-cells,leadingtomoreefficient activationofTcells,asIgG–antigenimmunecomplexesaretaken upmore efficiently than antigenalone via Fc␥R+expressedon antigen-presenting-cells;(2)IgM and IgG3enhancement ofthe antibodyresponsesthroughtheactivationofcomplement,which willthenlowerthethresholdforB-cellactivation[3].Mechanisms proposedforsuppressionofantibodyresponsesmaybe(1)forma- tionofIgG/antigenimmunecomplexeswhichmaybeeliminated byphagocyticcellsbeforetheycanactivatespecificBcells;(2)epi- topemaskingbyIgGwhichmaypreventB-cellsfromrecognising andrespondingtotheantigen[1–3].Theeffectofbaselineanti- bodyontheactivationofBcellsisnotwellunderstoodandyetis likelytohaveimportantimplicationsforthedesignofschedules forre-immunisation.
Whilsttheinhibitoryeffectofmaternalantibodyiswelldoc- umented[4,5]thesituation isless certainfor individuals given boosterdoses of vaccines following immunisation at an earlier time-point,becauseofconcernswithstatisticalmethodsandtheir interpretationinpublishedstudies.Manystudiesinhumanshave usedcorrelationorlinearregressiontoinvestigatehowtheabso- lute or relative increase in antibody concentration following a boosterdoseofvaccinerelatestopre-boosterantibodyconcentra- tion.Thesestudieshavereportedanegativecorrelationbetween baselineantibodyconcentrationandtheincreaseinantibodycon- centrationafteraboosterdoseofvaccine[6–9].Otherstudieshave investigatedtheratiobetweenpre-andpost-boosterantibodycon- centration (foldincrease) in individuals withdifferent baseline antibodyconcentrations(usinganarbitrarycut-off)andcommonly reportedhigherfoldincreasesinthegroupswithlowerpre-booster antibodytitre[6,7,10].
However,theuseofsimplecorrelationorlinearregressionto assesstherelationshipbetweenbaseline valuesand changes in valuehasbeencriticisedbystatisticiansformanyyears,sinceitis pronetobias[11,12].Theissueofconcernismathematicalcoupling betweenthetwovariables,i.e.thefactthatthechangescore(which iscorrelatedwiththebaselinevalue)isderivedfromasimpleequa- tionwhichincludesthebaselinevalueitself(changeiscalculated aspostvalueminusbaselinevalue).Thisisanimportantissue,for example,ifmeasurementerrorover-estimatestheobservedvalue ofpre-boosterantibodyconcentration(i.e.thereading ishigher thanitstruevalue),thenthereportedchangeinantibodyconcen- trationwillbelowerthanitstruevalue(sincethehighervalueof pre-boosterantibodyissubtractedincalculatingtheincreasein antibodyconcentration)[12].Ifmeasurementerrordecreasesthe observedvalueofpre-boosterantibodyconcentration,thentheval- uesofthepre-boosterantibodyconcentrationwillbelowerthanit shouldbeandthevaluesofthechangeinantibodyconcentration willbehigher.Thus,anegativecorrelationisaconsistentresult arisingsolelyfromthecombinationofmathematicalcouplingand measurementerrors[13].Measurementerrorhereisdeemedto includeshortterm fluctuationsin truevalues oftheantibodies beingmeasured,andanyrandomerrorsinherentinhandlingand storageofserumsamples,aswellastruemeasurementerrorofthe antibodylevels.Systematicerrorsattributabletostorageorhan- dlingmaynotaffectsresults.Forinstance,aconstantpercentage deteriorationinantibodycountsisequivalenttoaconstantvalue beingsubtractedfromallvaluesonthelogarithmicscale,sowill notaffectreportedassociationsbetweenbaselineandchangeson thelogscale(whichweareusinginthispaper).Themathematical
issuesapplysimilarlytocomparingthestrengthsofassociationsin sub-setswithdifferentdegreesofprotectionatbaseline.
Inresponsetotheseissues,statisticianshavedevelopedalterna- tivemethods[11–13,21],toassesstherelationshipbetweenchange andbaselinevalues,namelyOldham’smethodandBlomqvist’sfor- mula,whichhavedifferentmerits.Blomqvist’sformulaisbasedon knowledgeofwithinpersonfluctuationsinmeasurements(“error variance”).Thismaybeobtainedfromaseparatedataset,con- tainingrepeatedmeasurementsofantibodyconcentrationstaken overa shortperiod oftime. Intheabsenceofsuchdata,sensi- tivity analysesmaybeundertaken todeterminetheeffectof a plausiblerangeofestimatesofwithinpersonerrorvariances[11].
Oldham’smethodcorrelatesthechangebetweenpre-andpost- treatmentvalueswiththemeanofthepre-andpost-treatment values.Thismethodissaidtoreducetheeffectofmathematical couplingandbelesspronetobias.Undertherelativelyplausible assumptionsthatthemeasurementerror/shorttermfluctuations termsofpre-andpost-treatmentvaluesareindependentofeach other,and thatboth havethesame variance,then thismethod reducestheproblemsarisingfrommathematicalcoupling.These assumptionsareplausibleincomparisontotheassumptionofno measurementerror/shorttermfluctuationsrequired forvalidity of thestandard correlationor linearregression methods. How- ever,Oldham’smethodislikelytobesomewhatbiasedtowards apositiveassociationbetweenthechangeandmean(ofmedium termaveragesofbaselineand offollow-upconcentrations)val- ues. Sinceboth changeand meaninvolveadditionof follow-up values,anymeasurementerrorinthesevaluesinducesapositive bias.
Oldham’stechniquecanbeextendedtoallowconsiderationof additionalcovariatesorconfounders(e.g.,ageandgenderofstudy participants,variation inthe doseofvaccine antigens,studyin meta-analyses), ortoallowindividuals tobemeasuredvarying numbersoftimes.Thisrequiresuseofmulti-levelmodelling(MLM) techniques [12]. With correct model specification and without additionalcovariates,theresultsareequivalenttothoseobtained fromOldham’smethod.
Inthepresentstudy,we usedstandardcorrelationmethods, BlomqvistadjustmentmethodandanMLMversionofOldham’s methodtoassesstheeffectofpre-boosterMenC-IgGconcentra- tionontherelativeincreaseinMenC-IgGconcentrationfollowing booster immunisation in five studieswe previously conducted.
Since estimates of within person fluctuations in measurement (“error variance”) are not available, Blomqvist adjustment was basedonarangeofplausibleestimatesforthesevalues.
2. Materialsandmethods
Datawereobtainedfromfivestudiespreviouslyconductedin Oxford,UK[10,14–20].Inthesestudieschildrenwereprimedat2, 3,4monthsofagewitheitherserogroupCmeningococcalconju- gatevaccine(MenCV),oracombinedHaemophilusinfluenzaetype bandserogroupCNeisseriameningitidistetanustoxoidconjugate vaccine(Hib-MenC-TT), ora combined 9-valentpneumococcal- groupCmeningococcalconjugatevaccine(9vPnc-MenCV).They wereboostedat12monthsofagewitheitherMenCVormeningo- coccalA/Cpolysaccharidevaccine(MenA/C),or Hib-MenC-TT.A totalof6differentvaccinescheduleswereusedacrossthe5stud- ies(seeTable1)andthechildrenwereseparatedinto6groups accordingtotheirvaccineschedule.EachdoseofMenCVcontained 10gofN.meningitidisgroupColigosaccharide(MenC),however, theycontainedbetween12.5and25mcgofCRM197 [10,14–20].
TheyweremanufacturedeitherbyNovartisVaccinesandDiagnos- tics(studies1and3)[10MacLennan,2000#36,16],orbyWyeth Vaccines(studies2,4and5)[14,17,19].
Table1
Summarydescribingthestudiesfromwhichthedataweretakenforthepresentanalysis.
Studyno Studyname Groupofvaccineschedule Primingat2,3,4months Boostingat12months References
1 NovartisM14P5E1 1 MenCV MenCV [10]
2 WyethCombostudy 1 MenCV MenCV [14,15]
2 MenCV MenA/C
5 Pnc9-MenC MenCV
6 Pnc9-MenC MenA/C
3 NovartisMenCstudy 1 MenCV MenCV [16]
2 MenCV MenA/C
7a HepB MenCV
8a HepB MenA/C
4 WyethMenCstudy 1 MenCV MenCV [17,18]
2 MenCV MenA/C
7a HepB MenCV
8a HepB MenA/C
5 GSKHib-MenCstudy 4 Hib-MenC Hib-MenC [19,20]
3 MenCV Hib-MenC
aVaccineschedulenotusedinthepresentstudy.
EachdoseofMenA/Ccontained50gofMenC,however,only 1/5thoftheMenA/Cwasadministeredinstudies2and4[14,17], andthefulldoseofMenA/Cwasusedinstudy3[16].
ThecombinedHib-MenC-TTvaccinecontained5gofMenC, the combined Pnc9-MenC contained 10g of MenC. All the protein–polysaccharideconjugatevaccinesusedineachofthe6 studieshad CRM197 (a mutantof diphtheriatoxoid)as thecar- rierprotein,exceptthecombinedHib-MenC-TT,whichusedthe tetanustoxoidascarrierprotein.ThePnC9-MenCvaccinecontained 38.5mcgofCRM197[14].TheHib-MenC-TTcontained12.5mcgof tetanustoxoid[19].
TheconcentrationofIgGspecifictoMenCwasavailablebefore theboosterat12monthsofageandaftertheboosterat13months ofageforthechildrenparticipatinginallthestudies.Theconcen- trationsofIgGspecifictodiphtheriatoxoid(DT)beforethebooster wereavailableforallthestudiesexceptstudy4,andtheconcen- trationsoftetanustoxoid(TT)beforetheboosterwereavailablefor allthestudiesexceptstudies2and4.Foranyparticularstudy,the concentrationofeitherMenC,diphtheriaortetanus-specificIgG wasmeasuredinoneoffourlaboratories(OxfordVaccineGroup, UK;Novartis Vaccines(ex-ChironVaccines), Emeryville,Califor- nia;Wyeth,Rochester,NY;GlaxoSmithklineBiologicals,Belgium).
Theresultswerepooledandanalysedtoassesstheeffectofpre- boosterMenC-specificIgGconcentrationontherelativeincrease inMenC-IgGconcentrationafterthebooster.Theanalysiswasper- formedfirstusingthePearsons’scorrelationmethod,thenusing Blomqvist’s adjustment method to assess the plausible extend of bias to the standard correlation method, and finally using theMLMspecification ofOldham’s method[21]for eachgroup across 5 studies receiving 1 of 6 different vaccine schedules (seeTables1and2).
2.1. Blomqvistmethod
Foreachgroup,andindeedforeachstudywithingroups1and 2,Blomqvistadjustmentwasused.Thisadjuststheregressioncoef- ficient ofchanges onbaseline score (theunadjusted values are obtaineddirectlyfromourdatasets).Theadjustedregressioncoef- ficientsareconvertedtocorrelationcoefficientsusingthestandard formula;correlationofbaselineandchanges=(regressioncoeffi- cientofchangesonbaseline)SD(baseline)/SD (changes),where SD=standarddeviation.Thebaselineerrorvariancesusedwereas follows:theerrorvariancewasassumedtobe10%,and50%ofthe totalbaselinevarianceinthespecificgroupandstudyofinterest.
Theerrorvarianceatfollow-upwasassumedtobefirstlythesame asthatatbaseline,thenhalfthatatbaseline,andthenjust10%
ofthatatbaseline.Errorvariance inthechangescoresiscalcu- latedasthesumoferrorvariancesatbaselineandatfollow-up(by independenceoftheseerrorterms).Itisplausiblethattheerror varianceatfollow-upwillbesmallerthanatbaseline,sincethe meanvaluesaresubstantiallyhigher,soitmaybelesslikelythat theywillbeoutbyafactoroftwo(oranyotherchosenfactor), anditisthesefactordifferencesthatarerelevantwhenworking, ashere,onthelogarithmicscale.Forgroupswithmorethanone study,apooled correlationcoefficientwasderived bymeansof meta-analysis(inversevariancemethodonFishertransformation ofcorrelationcoefficients).
2.2. MLMversionofOldham’stechnique
For each group a MLM model was fitted using multilevel software(MLwiNversion2.0.01;http://www.cmm.bristol.ac.uk), where theoutcome ofthemodel wasthelogof theMenC-IgG concentration.Pre-andpost-treatment measureswerefittedas repeatedmeasureswithinsubject.Forgroupswithmorethanone study,a3-levelmodelwasfittedwithstudyatlevel3,subjects atlevel2,andtime(i.e.pre-andpost-boostercodedas-1and+1 respectively)atlevel1.A2-levelmodelwasusedforgroupswith onestudy,withsubjectsatlevel2andtime atlevel 1.Thecor- relationcoefficientwasderivedfromthecovariancebetweenthe randomslopeandrandominterceptofthemodel[21].Thisgives thecorrelationbetweenmean(ofpre-andpost-booster)valuesand changes.
Furthermore, in order to assess the role of carrier-specific antibodies onthe MenC-antibodies response tobooster immu- nisation, we assessed the correlations between the change in MenC-IgG concentration and the pre-booster diphtheria tox- oid/tetanus toxoid-IgG concentration (depending of the carrier proteincontainedinthevaccineusedforbooster)usingPearson’s correlationmethod,for thestudieswhere pre-boosterdiphthe- riatoxoid/tetanustoxoidIgG concentrationswereavailable.For groups with more than one study,a pooled correlation coeffi- cient wasderived bymeansof meta-analysis (inverse variance methodonFishertransformationofcorrelationcoefficients).Note thatthereisnomathematicalcouplingwhenlookingatcorrelation betweenbaselineprotein-carrierspecific-IgGconcentrationswith changesinMenC-IgGconcentrations,asthebaselinevariableisnot thesameasthevariableusedforthe“change”.
4156 G.Blanchard-Rohneretal./Vaccine30 (2012) 4153–4159
Table2 Analysisoftheeffectofpre-boosterMenCspecificIgGconcentration,andtherelativeincreaseinMenCspecificIgGconcentrationafterthebooster,usingunadjustedPearson’scorrelationcoefficients,theBlomqvistadjustment onthesecorrelations,andthenusingtheMLMtechnique. GroupPriming2,3,4months Boosting12monthsPre-boostMenC-IgG GMC(g/ml) (95%CI) Post-boost MenC-IgGGMC (g/ml) (95%CI) Men-CIgG GMR (95%CI) UnadjustedPearson’s correlation(preand change) (95%CI) Estimatedtruecorrelation(preandchange),usingBlomqvistmethodMLMcorrelation betweenmean(ofpre andpost)andchange (95%CI)10%errorvar, sameatf/u50%error var,same atf/u
10%errorvar, halfatf/u50%error var,halfat f/u
10%errorvar, 10%thatatf/u50%error var,10% thatatf/u 1(n=202)MenCVMenCV1.81(1.59,2.06)36(32,40)20(17,23)−0.58*(−0.48,−0.67)−0.57NA−0.55NA−0.53−0.08−0.15*(−0.29,−0.02) 2(n=130)MenCVMenA/C1.87(1.61,2.17)11(8.5,13)5.7(4.5,7.3)−0.5*(−0.36,−0.62)−0.51NA−0.48NA−0.46NA0.21*(0.04,0.39) 3(n=41)MenCVHib-MenC-TT1.11(0.8,1.55)3.7(2.9,4.7)3.3(2.6,4.3)−0.68*(−0.47,−0.82)−0.71NA−0.67NA−0.64−0.2−0.35*(−0.67,−0.03) 4(n=134)Hib-MenC-TT Hib-MenC-TT0.73(0.62,0.85)10(8.5,12)14(11,16)−0.58*(−0.46,−0.68)−0.56−0.42−0.55−0.3−0.54−0.25−0.02(−0.19,0.15) 5(n=44)9vPnc-MenCVMenCV0.51(0.41,0.63)6.5(4.6,9.2)13(8.9,19)−0.44*(−0.17,−0.65)−0.42−0.28−0.42−0.26−0.41−0.250.45*(0.13,0.76) 6(n=44)9VPnc- MenCVMenA/C0.4(0.32,0.49)2.2(1.7,2.9)5.6(4.4,7.3)−0.48*(−0.21,−0.68)−0.45−0.16−0.44−0.13−0.43−0.110.09(−0.19,0.38) ErrorvariancesforBlomqvistadjustmentaretakentobe10%and50%ofthetotalbaselinevariance,andatfollow-uparetakentoequaltheerrorvarianceatbaseline,ortobehalfofit,ortobe10%ofbaselineerrorvariance.NA, notavailable,missingvaluesforBlomqvistadjustedcorrelationsarearesultofthefollow-upvariancebeingtoosmalltosupportsuchlargeerrorvariancesasassumed. *Correlationthatarestatisticallysignificant.
AnalysiswascarriedoutusingStata(version10,StataCorp,USA) andMLwiN(version2.0.01,UniversityofBristol,UK).Dataofall analyseswerelogtransformedpriortoanalysestomeetthenormal distributionassumption.
3. Results
MenC-IgG antibody data were available from 646 children, whohadbeenimmunisedaccordingto1of6differentschedules (Table 1). The children were primed witheither MenCV, com- binedHib-MenC-TT,orcombined9vPnc-MenCV.Boostingwaswith either MenCV, MenA/C,or Hib-MenC-TT. Allgroups had a pre- booster MenC-geometric mean concentration(GMC) below the postulatedcorrelateofprotectionof2g/ml[22,23].However,the GMCofMenC-antibodyofgroups1,2,and3primedwithamono- valentMenCVwas≥1g/ml,incontrasttotheothergroupsprimed withacombinedvaccinecontainingtheMenCV.Followingbooster immunisation,allgroupshadasignificantincreaseinMenC-IgG concentration,andallthegroupshadaMenC-GMCabove2g/ml (Table2).
Using Pearson’s correlation method to assess the effect of pre-boosterMenC-IgGconcentrationontherelative increase in MenC-IgGconcentrationpost-booster,asignificantnegativecor- relation was observed for all the groups (Table 2). As already discussed,measurementerrorandshorttermfluctuationsinpre- boosterlevelscouldcause,orcontributetosuchassociations.The Blomqvistadjustmentshowshowmuchofaninfluencethisbias mayplausiblyhave,byshowingestimatedcorrelationcoefficients betweenunderlyingtruevaluesatbaselineandchanges(i.e.taken frombaselineandfollow-upvaluesaveragedovershorttermfluc- tuationswithoutmeasurementserror,Table2).Blomqvistadjusted correlationcoefficientsremainnegativeinmostcasesexceptwhen theerrorvarianceishigh.Forexampleforgroup6,primedwith 9vPnc-MenCVandboostedwithMenA/C,wheretheadjustedcor- relationis−0.16,when50%ofthebaselinevarianceisattributable toerror,evenassumingthesameerrorvarianceatfollow-up.
UsingtheMLMtechniquetoassesstherelationshipbetween mean(ofpre-andpost-booster)MenC-IgGconcentrationandthe relative increase in MenC-IgG concentration post-booster, con- trastingresultswereobservedwitheitherapositive,negativeor nocorrelation(Table2).
Toassesstherelationshipbetweencarrier-proteinspecificanti- bodyconcentrationsandtheincrease inMenC-specificantibody concentrationsPearsons’scorrelationcoefficientwasused.Note thatinthiscasetheuseofPearsons’scorrelationisvalidsincethe variablesconcernedarenotmathematicallycoupled.Anegative correlationwasobservedwhena protein–polysaccharideconju- gatevaccine(MenCVorHib-MenC-TT)wasusedforbooster.This wassignificantfor3outofthe4suchgroups(Table3).Incontrast, therewasnocorrelationwhentheplain-polysaccharideMenA/C vaccinewasusedasbooster(Table2).
4. Discussion
Theeffectofpre-existingantigen-specificserumantibodyon responsetosubsequentdosesofvaccineisnotwellunderstood.
Studying this phenomenon in humans, outside the settings of transplacentalantibodytransferandclinicalsituationswhereanti- bodiesareadministeredpassively(e.g.hepatitisBimmunoglobulin inindividualsexposedtoaninfectioussource),islimitedtoobser- vational studies usingcorrelation or regression analysis.In the currentstudyusingstandardcorrelationmethodstheincreasein MenC-antibodyconcentrationwasconsistentlynegativelycorre- latedwithpre-boosterMenC-antibodyconcentrationaspreviously described [6–9]. Negative correlations, in other disease areas,
Table3
CorrelationbetweenthechangeinMenC-IgGconcentrationandthepre-boostercarrierprotein-IgGconcentrationusingPearson’scorrelationmethod.
GroupPrimingat2,3,4months Boostingat12months Studyno. No.ofparti-cipants Correlationbetweenpre-boost carrierproteinaIgGandchangein MenCIgGonboosting(95%CI)
p-Value
1 MenCV MenCV 1,2,3 110 −0.43(−0.58,−0.26) <0.0001
2 MenCV MenA/C 2,3 97 0.06(−0.14,0.26) 0.57
3 MenCV Hib-MenC-TT 5 41 −0.21(−0.49,0.1) 0.19
4 Hib-MenC-TT Hib-MenC-TT 5 132 −0.18(−0.34,0) 0.04
5 9vPnC-MenCV MenCV 2 35 −0.47(−0.7,−0.17) 0.004
6 9vPnC-MenCV MenA/C 2 38 −0.04(−0.36,0.28) 0.8
aDiphtheriatoxoid-IgGwasusedforallthegroupsexceptgroups3and4wheretetanustoxoidwasusedbecausetheHib-MenC-TTcontainstetanustoxoidascarrier protein.
betweenbaselineandchangeshavebeenshowntoresultfromthe combinedeffectofmathematicalcouplingandmeasurementerror [24,25],andsocannotbeinterpretableinaclinicallymeaningful way.UseofBlomqvistadjustment,onarangeofplausibleerror variances,suggeststhatthetrueunderlyingcorrelationbetween baselineandchangesremainsnegativeinthecasesofsmallerror variance,asitcanbeexpectedassumingthatthereisasmallcoef- ficientofvariationintheELISAassayused[26,27]andassuming thatthereislittledaytodayfluctuationinMenC-IgGconcentra- tions[10].In contrast,usingtheMLMtechnique,toaccountfor themathematicalcouplingofthepre-andpost-boosterantibody concentration,aconsistentnegativecorrelationwasnotobserved betweenthemeanMenC-specificantibodyconcentrationandthe MenC-specificantibodyincreasepost-booster.However,theuseof theMLMmethodappearsnottobeappropriateasitmayintroduce apositivebias,sincethefollow-upvaluesareaddedincalculating bothbaselineandchanges,soanymeasurementerrorinfollow-up terminducesapositivebias.Furthermore,itdoesnotdirectlymea- surethecorrelationbetweenbaselineandchange(ratherituses meanandchange).Thesedatasuggestthattheinhibitoryeffect ofbaselineantibodyontheimmuneresponsetoboostermaybe accurateassumingthatthereisalowerrorvariance(duetosmall measurementerrorduetotheELISAassayand littledaytoday fluctuationsinmeasurements)[10,26,27].
TheuseofOldham’smethodassociatedwithMLMtoaccount formathematicalcouplingbetweenbaselinevaluesandchangesis supportedbyTuetal.fromaseriesofpublicationsmostlyonden- talliterature.However,thismethodhasalsobeencriticisedasitis basedonseveralassumptions(suchassameerrorvarianceatbase- lineandatfollow-up)whichcannotbeverified.Furthermore,this methoddoesnotdirectlymeasurethecorrelationbetweenbase- lineandchanges,whereasthatisthecorrelationwhichisclinically relevant(ratherthantheOldham’scorrelationbetweenmeanand changes).OtherssupporttheuseofBlomqvist[11],althoughaddi- tionaldata(repeatedmeasurementstakenshortintervalsapart)are neededtoapplythistechniquedirectly,whichisnotreallyfeasi- blewhendoingvaccinetrialsinyoungchildrenwithlimitedblood draws.However,we haveusedthemethodsuggestedbyHayes etal.(1988)amongothers,whichiseffectivelyundertakingasen- sitivityanalysis,toestimateerrorvariance,accordingtoarangeof plausibleestimates.Asdiscussedearlier,onlyalowerrorvariance islikelytooccurinthestudiesreportedinthepresentmanuscript.
Ourstudyislimitedbytheheterogeneityofthedifferentstud- iesused.Indeed,sinceavarietyofotherfactorsvariedamongstthe infantsstudied[e.g.thedoseofMenCusedinprimingandboosting, theroleofcombinationversusmonovalentvaccineinprimingor boosting,thegenerallevelofpre-boostantibody(highervslower)]
itwasnotpossibletoinferwhichoneswereindependentlyimpor- tantduetothenumberofgroupsavailableforstudyinrelationto thenumberofpossiblevariables.Also,theantigen-specificIgGcon- centrationsweremeasuredindifferentlaboratories,andalthough laboratoriesderivedtheiranti-polysaccharideIgGprotocolsfrom twopublications[26,28],theassayspooled,analysedandcompared inthisstudyhavenotbeencomparedtoeachotherdirectly,which shouldbeacknowledgedasa limitationofthis study.However, themethodusedaccountsforthisheterogeneity(causedbyassay variationsorinfantselection)betweenthedifferentstudiesused.
There are a number of factors that might explain why pre-immunisation antibody concentrations may have a weaker inhibitory effect on vaccine immunogenicity than previously reported.First, thestatistical methodsused inprevious studies didnotaccountfortheproblemsofcouplingwhichareinherent in studieswherea variableis measuredinthesame individual pre-andpost-intervention(asdiscussedabove).Second,antibody mediatedinhibitionmaynotbeasignificantimmunologicalphe- nomenoninthecontextofboosterimmunisationofinfantsorat leastitmayonlybesignificantathigherconcentrationsofpre- immunisationantibodythantheonesobservedintheinfantsof thepresentstudy.Inseveralreviewsofanimaldata,ithasbeen reportedthattheantibody-feedbackregulationonBcellresponses toimmunisationcouldbeeitherpositiveornegative[1–3].Finally, analternativepossibilityisthatevenifantibodymediatedinhibi- tionoccurs,thepre-immunisationantibodycorrelateswithsome otherimmunologicalparameter(e.g.immunologicalmemory)that ismoresignificantinenhancingtheimmuneresponse thanthe antibodyisatdirectlyinhibitingit.Itisthoughtthatserumanti- bodyismaintainedbythedifferentiationofmemoryBcellsinto plasmacells.Iftheantibodylevelsatbaselinereflectthesizeofthe memoryBcellpoolavailable,thenonemightexpectgreatersec- ondaryantibodyresponsestobeassociatedwithhigherlevelsof pre-existingantibody.
Astrikingfindingof thecurrentstudywasaconsistentneg- ative correlation between the carrier protein-specific baseline antibodyconcentrationandtheincreaseinMenC-specificantibody concentrationfollowingboosterimmunisationwiththeprotein- polysaccharideconjugatevaccine,whichwassignificantin3out of the4 groups tested. Note that theonly group (group 3)for whichthecorrelationwasnotsignificantwasagroupwitharela- tivelysmallsamplesize.Incontrast,therewasnocorrelationwhen theplain-polysaccharideMenA/Cvaccinewasusedasbooster.For variableswhichaprioriareunrelatedstandardcorrelationmeth- odscanbeused,sincethereisnomathematicalcouplingpresent.
“Carrierinducedepitopesuppression”describestheeffectofpre- existing immunitytoa carrier-protein, onresponse toa linked
4158 G.Blanchard-Rohneretal./Vaccine30 (2012) 4153–4159
haptenicantigen(inthiscasethepolysaccharide)[14,29].Avariety ofmechanismshavebeendescribedinvolvingbothcarrier-protein specificBandT-cellsinadditiontopre-existingserumantibody [30]. Preformed Bcells and T cells specific for the carrierpro- teinscouldcompeteforantigencapturewithBcellsspecificfor thepolysaccharide,thusinhibitingtheimmuneresponse tothe polysaccharidecomponentoftheconjugatevaccine[30–32].These findingssuggestthatthelevelofcarrier-specificantibodymight correlatewiththelevelofcarrier-specificBcells,whichmaycom- petewith polysaccharide-specificBcells for thecaptureof the MenCV. An alternative explanationfor this could bethat anti- bodytothecarrierproteinprovidesmoreeffectivesterichindrance oftheinteractionofvaccineantigenandantigen-specificB-cells thantheanti-polysaccharideantibody[30].Thiscouldbedueto eitheroftheantigenqualityandquantityorthatoftheantibody.
Anintriguingpossibilityisthatpre-existingcarrier-specificanti- body may bea more potent inhibitor of T-cell responses than B-cells.Fromthispointofviewitwouldbeusefultolookatanti- bodymediatedinhibitionduetoavaccineantigencontainingonly T-dependentconstituent antigens. However, another studyhas reportedthat high pre-booster concentrationof carrier-specific antibodycorrelatedpositivelywiththeincreaseinantibodyspe- cificforthepolysaccharideantigenfollowingimmunisationwith a H. influenzae type b conjugate vaccine in adults [33]. It was suggestedbytheinvestigatorsof thisstudythatcarrier-specific antibodiescorrelatedwiththenumbersofcarrier-specifichelperT cells,whichinturncouldincreasetheBcellresponseagainstthe polysaccharidecomponentoftheconjugatevaccine.Therefore,the phenomenonofcarrier-inducedepitopesuppressionmaydepend onseveralfactors. Forexample, thedoseof thecarriermaybe important.Itwasobservedinamurinestudythatlow-dosecarrier primingimprovedsubsequentresponsetopolysaccharidescon- jugatedtothesamecarrier,whilehighdoseprimingsuppressed theresponse [34].Similarly,it wasreportedthat thekey factor for the inhibitory effect of maternal antibodies on infant vac- cineresponseswastheratioofantibodytoimmunisingantigen [4].Inthepresentstudy,theconcentrationofCRM197wasquite highforall thevaccines (MenCV and9vPnC-MenCV), however, thedoseoftetanustoxoidusedintheHib-MenC-TTvaccinewas quitelow.
Arecommendationforfurtherresearchisthat themeasure- mentsofMenC-IgGandprotein-carrierspecific-IgGaretakenat leasttwiceatashorttimeintervalapart,certainlyaroundbaseline andpreferablyalsoatfollow-up,sothatthevarianceofthemea- surementerror/shorttermfluctuationsinthesemeasurementscan befounddirectly.ThiswouldallowBlomqvist’smethodtobeused directlytoadjustthecorrelationsbetweenpre-andpost-MenC-IgG measurements,andgivefurtherinsightintotheresultspresented here.Itwouldalsoallowforadjustmentstobemadetothecorrela- tioncoefficientsbetweenbaselineprotein-carrierspecific-IgGand changesinMenC-IgG.Itwouldalsobedesirableinfuturevaccine trialsingeneraltobuildrepeatedmeasurementsofantibodiesinto theprotocol,atleastforasubsetofpatients,sothattheseissues canbemorefullyassessed.
Inconclusion,wehavedemonstratedthatunderstandingthe effectof pre-existent antibody onresponse toimmunisation is dependentonusing statisticalmethods that account for math- ematical coupling.Using Blomqvist method of adjustment and assumingthat thereis low errorvariance,we haveshown that theinhibitoryeffectofpre-existingMenCantibodyconcentration ontheMenC-antibodyresponsetoboosterimmunisationreported previouslymightbeaccurate.Incontrast,therewasaconsistent inhibitionoftheproteincarrier-specificantibodypre-boosteron theMenC-antibodyincrease post-boosterwiththeMenCV sug- gestiveofthefactthateither(i)serumantibodyisinhibitorythe effectofimmunological memoryis excludedor(ii)that carrier
proteinantibodiesaremorelikelytomediateinhibitionthrough mechanismssuchasdiversionoralterationsinT-cellhelp.These findingsshouldbefurtherexploredastheyhaveimplicationsfor theoptimaltimingofdosesofconjugatevaccinesandtheirrela- tionshipto boosterdosesof vaccines containingcarrier-protein epitopes.
Conflictofinterests
AJPhasconductedclinicaltrialsonbehalfofOxford Univer- sity, sponsored by Wyeth Vaccines, GlaxoSmithKline Vaccines, SanofiPasteur,SanofiPasteurMSDandNovartisVaccines,butdoes notaccept anypersonalpayments fromvaccinemanufacturers.
Industry-sourcedhonorariaforlecturing,writingor consultancy andtravelexpensesforattendanceatscientificmeetingsarepaidto aneducational/administrativefundheldbytheDepartmentofPae- diatrics,UniversityofOxford.MDShasreceivedassistancetoattend scientificmeetingsfromWyethVaccines,NovartisVaccinesand GlaxoSmithKlineVaccinesandhashadtravelandaccommodation expensespaidbyNovartisVaccineswhileworkingincollaboration withNovartisVaccinesinSiena,Italy.
Acknowledgements
Wewouldliketothankthefollowingcompanies(Novartis,GSK and Wyeth)for allowingustousetheirdata,and inparticular GSKforveryusefulcommentsonthemanuscript.G.B.-R.wassup- portedbytheSwissNationalScienceFoundation.DFKandMDS arefundedbytheOxfordPartnershipComprehensiveBiomedical Research CentrewithfundingfromtheDepartment of Health’s NIHR Biomedical Research Centres funding scheme. The views expressedinthispublicationarethoseoftheauthorsandnotnec- essarilythoseoftheDepartmentofHealth.A.J.P.isaJennerInstitute InvestigatorandJamesMartinFellow.
References
[1]HeymanB.Regulationofantibodyresponsesviaantibodies,complement,and Fcreceptors.AnnualReviewofImmunology2000;18:709–37.
[2] Heyman B. Feedback regulation by IgG antibodies. Immunology Letters 2003;88(2):157–61(August5).
[3]HjelmF,CarlssonF,GetahunA,HeymanB.Antibody-mediatedregulationofthe immuneresponse.ScandinavianJournalofImmunology2006;64(September (3)):177–84.
[4]SiegristCA.Mechanismsbywhichmaternalantibodiesinfluenceinfantvaccine responses:reviewofhypothesesanddefinitionofmaindeterminants.Vaccine 2003;21(24):3406–12(July28).
[5]SiegristCA,BarriosC,MartinezX,BrandtC,BerneyM,CordovaM,etal.Influence ofmaternalantibodiesonvaccineresponses:inhibitionofantibodybutnotT cellresponsesallowssuccessfulearlyprime-booststrategiesinmice.European JournalofImmunology1998;28(December(12)):4138–48.
[6]Danilova E,ShiryayevA,KristoffersenEK,SjursenH.Attenuated immune responsetotetanustoxoidinyounghealthymenprotectedagainsttetanus.
Vaccine2005;23(42):4980–3(October10).
[7]DanilovaE,ShiryayevA,SkogenV,KristoffersenEK,SjursenH.Short-term boostereffectofdiphtheriatoxoidininitiallylong-termprotectedindividuals.
Vaccine2005;23(12):1446–50(February10).
[8]RonneT,ValentelisR,TarumS,GriskevicaA,WachmannCH,AggerbeckH,etal.
ImmuneresponsetodiphtheriaboostervaccineintheBalticstates.TheJournal ofInfectiousDiseases2000;181(February(Suppl.1)):S213–9.
[9]NicolayU,GirgsdiesOE,BanzhoffA,HundtE,JilgW.Diphtheriaboostervacci- nation:oneortwoinjections?Vaccine1999;17(18):2223–8(May4).
[10]BlanchardRohnerG,SnapeMD,KellyDF,JohnT,MorantA,YuLM,etal.The magnitudeoftheantibodyandmemoryBcellresponsesduringprimingwitha protein–polysaccharideconjugatevaccineinhumaninfantsisassociatedwith thepersistenceofantibodyandtheintensityofboosterresponse.TheJournal ofImmunology2008;180(4):2165–73(Feb15).
[11]HayesRJ.Methodsforassessingwhetherchangedependsoninitialvalue.
StatisticsinMedicine1988;7(September(9)):915–27.
[12]Tu YK,GilthorpeMS. Revisitingthe relationbetween changeandinitial value: areviewand evaluation.Statisticsin Medicine2007;26(2):443–57 (January30).
[13] OldhamPD.Anoteontheanalysisofrepeatedmeasurementsofthesame subjects.JournalofChronicDiseases1962;October(15):969–77.
[14]ButteryJP,RiddellA,McVernonJ,ChantlerT,LaneL,Bowen-MorrisJ,etal.
Immunogenicityandsafetyofacombinationpneumococcal-meningococcal vaccineininfants:arandomizedcontrolledtrial.JournaloftheAmericanMed- icalAssociation2005;293(14):1751–8(April13).
[15]RiddellA,ButteryJP,McVernonJ,ChantlerT,LaneL,Bowen-MorrisJ,etal.A randomizedstudycomparingthesafetyandimmunogenicityofaconjugate vaccinecombinationcontainingmeningococcalgroupCandpneumococcal capsularpolysaccharide–CRM197withameningococcalgroupCconjugate vaccineinhealthyinfants:challengephase.Vaccine2007;25(19):3906–12 (May10).
[16]MacLennanJM,ShackleyF,HeathPT,DeeksJJ,FlamankC,HerbertM,etal.
Safety,immunogenicity,andinductionofimmunologicmemorybyaserogroup C meningococcalconjugate vaccine in infants: a randomized controlled trial.JournaloftheAmericanMedicalAssociation2000;283(21):2795–801 (June7).
[17] EnglishM,MacLennanJM,Bowen-MorrisJM,DeeksJ,BoardmanM,Brown K,etal.Arandomised,double-blind,controlledtrialoftheimmunogenicity andtolerabilityofameningococcalgroupCconjugatevaccineinyoungBritish infants.Vaccine2000;19(9–10):1232–8(December8).
[18] SnapeMD,MaclennanJM,LockhartS,EnglishM,YuLM,MoxonRE,etal.
DemonstrationofimmunologicmemoryusingserogroupCmeningococcalgly- coconjugatevaccine.ThePediatricInfectiousDiseaseJournal2009;28(February (2)):92–7.
[19]PaceD,SnapeM,WestcarS,OluwalanaC,YuLM,BeggN,etal.Anovelcom- binedHib-MenC-TTglycoconjugatevaccineasaboosterdosefortoddlers:a phase3openrandomisedcontrolledtrial.ArchivesofDiseaseinChildhood 2008;93(November(11)):963–70.
[20]PaceD, SnapeM,WestcarS,HamalubaM,YuLM,BeggN,etal. Anew combination haemophilus influenzae type B and Neisseria meningitidis serogroupC-tetanustoxoidconjugatevaccineforprimaryimmunizationof infants.ThePediatric InfectiousDiseaseJournal2007;26(November(11)):
1057–9.
[21]BlanceA,TuYK,GilthorpeMS.Amultilevelmodellingsolutiontomathe- maticalcoupling.StatisticalMethodsinMedicalResearch2005;14(December (6)):553–65.
[22] PeltolaH.Meningococcalvaccines:currentstatusandfuturepossibilities.
Drugs1998;55(March(3)):347–66.
[23]GoldR, Lepow ML,GoldschneiderI, DraperTF,GotshlichEC.Kineticsof antibodyproductiontogroupAandgroupCmeningococcalpolysaccharide vaccinesadministeredduringthefirstsixyearsoflife:prospectsforrou- tineimmunizationofinfantsandchildren.TheJournalofInfectiousDiseases 1979;140(November(5)):690–7.
[24]TuYK,MaddickIH,GriffithsGS,GilthorpeMS.Mathematicalcouplingcan underminethestatisticalassessmentofclinicalresearch:illustrationfromthe treatmentofguidedtissueregeneration.JournalofDentistry2004;32(February (2)):133–42.
[25]TuYK,GilthorpeMS,GriffithsGS.Isreductionofpocketprobingdepthcorre- latedwiththebaselinevalueorisitmathematicalcoupling?JournalofDental Research2002;81(October(10)):722–6.
[26] Gheesling LL, Carlone GM, Pais LB, Holder PF, Maslanka SE, Plikaytis BD, et al. Multicenter comparison of Neisseria meningitidis serogroup C anti-capsular polysaccharideantibody levels measured by a standard- izedenzyme-linkedimmunosorbentassay.JournalofClinicalMicrobiology 1994;32(June(6)):1475–82.
[27]ElieCM,HolderPK,Romero-SteinerS,CarloneGM.Assignmentofadditional anticapsularantibodyconcentrationstotheNeisseriameningitidisgroupA,C, Y,andW-135meningococcalstandardreferenceserumCDC1992.Clinicaland DiagnosticLaboratoryImmunology2002;9(May(3)):725–6.
[28]GranoffDM,MaslankaSE,CarloneGM,PlikaytisBD,SantosGF,MokatrinA,etal.
Amodifiedenzyme-linkedimmunosorbentassayformeasurementofantibody responsestomeningococcalCpolysaccharidethatcorrelatewithbacterici- dalresponses.ClinicalandDiagnosticLaboratoryImmunology1998;5(July (4)):479–85.
[29]DaganR,EskolaJ,LeclercC,LeroyO.Reducedresponsetomultiplevaccines sharingcommonproteinepitopesthatareadministeredsimultaneouslyto infants.InfectionandImmunity1998;66(May(5)):2093–8.
[30]DaganR,PoolmanJ,SiegristCA.Glycoconjugatevaccinesandimmuneinter- ference:areview.Vaccine2010;28(34):5513–23.
[31]BaringtonT,SkettrupM,JuulL,HeilmannC.Non-epitope-specificsuppression oftheantibodyresponsetoHaemophilusinfluenzaetypebconjugatevac- cinesbypreimmunizationwithvaccinecomponents.InfectionandImmunity 1993;61(February(2)):432–8.
[32]InselRA.Potentialalterationsinimmunogenicitybycombiningorsimultane- ouslyadministeringvaccinecomponents.AnnalsoftheNewYorkAcademyof Sciences1995;754:35–47(May31).
[33]BaringtonT,KristensenK,HenrichsenJ,HeilmannC.Influenceofprevac- cinationimmunityonthehumanB-lymphocyteresponsetoaHaemophilus influenzaetypebconjugatevaccine.InfectionandImmunity1991;59(March (3)):1057–64.
[34]DaumRS,ZenkoCE,GivenGZ,BallancoGA,ParikhH,GerminoK.Magnitudeof interferenceafterdiphtheria-tetanustoxoids-acellularpertussis/Haemophilus influenzae type b capsular polysaccharide-tetanus vaccination is related tothe numberof dosesadministered.TheJournal ofInfectiousDiseases 2001;184(10):1293–9(November15).