• No results found

Factors affecting the cost effectiveness of on farm culture prior to the treatment of clinical mastitis in dairy cows

N/A
N/A
Protected

Academic year: 2020

Share "Factors affecting the cost effectiveness of on farm culture prior to the treatment of clinical mastitis in dairy cows"

Copied!
9
0
0

Loading.... (view fulltext now)

Full text

(1)

Contents lists available atScienceDirect

Preventive

Veterinary

Medicine

j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / p r e v e t m e d

Factors

affecting

the

cost-effectiveness

of

on-farm

culture

prior

to

the

treatment

of

clinical

mastitis

in

dairy

cows

P.M.

Down

a,∗

,

A.J.

Bradley

b

,

J.E.

Breen

a,b

,

M.J.

Green

a

aUniversityofNottingham,SchoolofVeterinaryScienceandMedicine,SuttonBoningtonCampus,SuttonBonington,LoughboroughLE125RD,United

Kingdom

bQualityMilkManagementServicesLtd,CedarBarn,EastonHill,Easton,WellsBA51DU,UnitedKingdom

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received30March2017

Receivedinrevisedform6July2017 Accepted7July2017

Keywords:

Mastitis On-farmculture Cost-effectiveness Treatment

Probabilisticsensitivityanalysis

a

b

s

t

r

a

c

t

Theobjectiveofthisstudywastouseprobabilisticsensitivityanalysistoevaluatethecost-effectivenessof usinganon-farmculture(OFC)approachtothetreatmentofclinicalmastitisindairycowsandcompare thistoa‘standard’treatmentapproach.Aspecificaimwastoidentifytheherdcircumstancesunder whichanOFCapproachwouldbemostlikelytobecost-effective.AstochasticMonteCarlomodelwas developedtosimulate5000casesofclinicalmastitisatthecowlevelandtocalculatetheassociatedcosts simultaneouslywhentreatedaccordingto2differenttreatmentprotocols;i)a’conventional’approach (3tubesofintramammaryantibiotic)andii)anOFCprogramme,wherebycowsaretreatedaccordingto theresultsofOFC.ModelparametersweretakenfromrecentpeerreviewedliteratureontheuseofOFC priortotreatmentofclinicalmastitis.Spearmanrankcorrelationcoefficientswereusedtoevaluatethe relationshipsbetweenmodelinputvaluesandtheestimateddifferenceincostbetweenthestandardand OFCtreatmentprotocols.Thesimulationanalysesrevealedthatboththedifferenceinthebacteriological curerateduetoadelayintreatmentwhenusingOFCandtheproportionofGram-positivecasesthatoccur onadairyunitwouldhaveafundamentalimpactonwhetherOFCwouldbecost-effective.Theresultsof thisstudyillustratedthatanOFCapproachforthetreatmentofclinicalmastitiswouldprobablynotbe cost-effectiveinmanycircumstances,inparticular,notthoseinwhichGram-positivepathogenswere responsibleformorethan20%ofallclinicalcases.Theresultshighlightanethicaldilemmasurrounding reduceduseofantimicrobialsforclinicalmastitissinceitmaybeassociatedwithfinanciallossesand poorercowwelfareinmanyinstances.

©2017TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).

1. Introduction

Mastitisisoneofthemostprevalentandcostlydiseases affect-ingdairycowsworldwide.Thecostofclinicalmastitisismadeupof ‘direct’costs,suchasthecostofdrugs,discardedmilkandincreased labour and ‘indirect’ costs,such as reduced future production, increasedcullingandincreasedtheriskofdiseasetransmissionto herdmates.Theoverallcostofacaseofclinicalmastitishasbeen showntobehighlyvariable(Huijpsetal.,2008;Greenetal.,2009) andmostinfluencedby‘indirect’costs(KossaibatiandEsslemont,

2000;Huijpsetal.,2008;Downetal.,2013).

Notonlyismastitisimportantintermsoftheeconomics,but thetreatmentandpreventionofmastitisiswidelyreportedasthe mostcommonreasonforantimicrobialdruguseondairyfarms

∗Correspondingauthor.

E-mailaddress:peter.down@nottingham.ac.uk(P.M.Down).

(PolandRuegg,2007;Thomsonetal.,2008;Gonzálezetal.,2010).

Thereisincreasingpressureontheagriculturalsectortoreduce antimicrobialdrugusageduetofearsoverantimicrobialresistance (AMR)(O’Neill,2015)andthewayinwhichantimicrobialdrugs areappliedwithrespecttothetreatmentofmastitisis,therefore,a sensibletarget.Conventionally,allcasesofclinicalmastitiswould receiveacourseofantimicrobialagentsbutanalternativeapproach istheselectivetreatmentofcasesaccordingtotheresultsofan on-farmculture(OFC)system.WiththeOFCsystem,onlynon-severe casesthatyieldaGram-positiveormixed-culturearetreatedwith antimicrobialdrugsresultinginmanycasesofclinicalmastitisnot beingtreatedat all(Lago etal.,2011a).Thiswasdemonstrated recentlyinastudyperformedin8herdsbasedinMinnesota, Wis-consinandOntario,whichreportedthat51%ofcowsenrolledin theOFCgroupreceivedantimicrobialdrugsasopposed to100% ofthecowsenrolledintheconventionalgroup.Thesamestudy reportednostatisticaldifferencesbetweenthetwogroupswith respecttothebacteriologicalcurerisk,thetimetakentoclinical

http://dx.doi.org/10.1016/j.prevetmed.2017.07.006

(2)

92 P.M.Downetal./PreventiveVeterinaryMedicine145(2017)91–99

cure,newintramammaryinfectionrisk,treatmentfailureriskor riskofremovalfromtheherdwithin21days(Lagoetal.,2011a).

WhileOFCappearstobeeffectiveinreducingantimicrobialdrug usage;littleisknownaboutfactorsinfluencingtheoverall cost-effectivenessofthisapproachand,therefore,thetypesofherdsin whichitismostlikelytobecost-effective.Whenperformingsucha cost-effectivenessanalysis,thereareinevitablymultiplesourcesof evidenceforparameterestimatesandadegreeofuncertainty sur-roundingtheirtruevalue(Adesetal.,2006).Fordecision-making purposes,itisimportantthatcost-effectivenessmodelsareableto incorporatemultiplesourcesofevidenceandreflectuncertaintyin themodeloutputs(Claxtonetal.,2005;BaboMartinsandRushton, 2014).An approachtothis now widelyreported inthehuman healthliterature(Briggsetal.,2002;Brownetal.,2006)and increas-inglyintheveterinaryliterature(Downetal.,2013;Hudsonetal.,

2014,2015)is probabilistic sensitivityanalysis(PSA).The main

featureofthis techniqueisthat allinputparametersare speci-fiedasfullprobabilitydistributions,ratherthanpointestimates, torepresenttheuncertaintysurroundingtheirvalues.This param-eteruncertaintyisthenpropagatedthroughthecost-effectiveness modelsothattheresultingimprecisionisreflectedinmodel out-putsandinferencesmade(Briggsetal.,2002).

Thepurposeofthisresearchwastouseprobabilisticsensitivity analysistoinvestigatethemainfactorsthat influencethe cost-effectivenessofanOFCapproachtotreatingclinicalmastitis.The modelusedwasanadaptationofonereportedpreviously(Down etal.,2013),withtheadditionofOFC-specificparametersbasedon previousresearch(Lagoetal.,2011a).Aspecificaimwasto iden-tifytheherdcircumstancesunderwhichanOFCapproachwould bemostlikelytobecost-effective.

2. Materialsandmethods

2.1. Modelstructure

AstochasticMonte Carlomodel wasdeveloped using Open-BUGS3.2.2software(Thomasetal.,2006).Themodel wasused tosimulateacaseofclinicalmastitisatthecowlevelandto cal-culatetheassociatedcostssimultaneouslywhentreatedaccording to2differenttreatmentprotocols;i)a‘conventional’approach(3 tubesofintramammaryantibiotic)andii)anOFCprogrammeas describedbyLagoetal.(2011a)inwhichmilksamplestakenfrom cowswithCMwereculturedon-farmusingtheMinnesotaEasy CultureSystem(UniversityofMinnesota,St.Paul).Thison-farm milkculturesystemconsistedofabi-plate,whichisaPetridish withtwodifferenttypesofagar:MacConkeyagarononehalfthat selectivelygrowsGram-negativebacteriaandFactormediaonthe otherhalfoftheplatethatselectivelygrowsGram-positive bacte-ria.Theplatewasplacedinanon-farmincubatorandincubated atapproximately37◦Cfor18–24h.Thenextday,theplatewas examinedforbacterialgrowthandinterpretedbyherdpersonnel andifbacteriadidnotgrow,theplatewasreturnedtotheincubator andre-readapproximately18–24hlater.Resultsforeachsample platewererecordedas(1)Gram-positive,whenbacteriagrewonly intheFactoragarmediaofthebi-plate;(2)Gram-negative,when bacteriagrewonlyintheMacConkeyagarmediaofthebi-plate; (3)nogrowth,when bacteriadidnotgrowoneithermedia;or (4)mixedinfectionwhenbacteriagrewonbothmedia.Quarters fromwhichGram-positivebacteriawereisolatedorhadamixed infectionreceivedintramammaryantibiotictreatmentandifthe on-farmmilkcultureresultwasGram-negativeornogrowth,the quarterdidnotreceiveintramammarytherapy.

Thegeneralmodelstructureandassumptionswereconsistent irrespectiveofthetreatmentprotocolappliedandwasasdescribed previouslybyDownetal.(2013)(Fig.1).Aninitialcaseofclinical

mastitis(CM1)couldeither;i)curebacteriologically,ii)cure clin-icallybutremainsubclinicallyinfected,oriii)failtocure(either clinicallyorbacteriologically).IfCM1failedtocure,thenarepeat treatment(sameasinitialtreatment)wouldbeadministeredand thesamethreeoutcomespermitted.IfCM1curedbacteriologically, thenthecowcouldeitherend thelactationorbeculledbefore theendoflactation.IfCM1curedclinicallybutnot bacteriologi-cally,thenitcouldeitherendthelactation,beculledbeforethe endoflactation,orhavearepeatepisodeofclinicalmastitis(CM2). CM2wouldbetreatedaccordingtothesameprotocolasCM1and wouldthenfollow thesamepossibleoutcomesasCM1.Athird clinicalrecurrencewaspermittedforsubclinicallyinfectedcows (CM3)whichwereagaintreatedinthesamewayasCM1andCM2. IfthecowremainedsubclinicallyinfectedafterCM3orfailedto cureclinically,thenthecowwouldbeculledbeforetheendof lac-tation.IfthecowcuredbacteriologicallyafterCM3,thenitcould eitherfinishthelactationorbeculledbeforetheendoflactation.

Ariskoftransmissionparameterwasincludedfromcowsthat remainedsubclinicallyinfectedafterCM1 andCM2.This repre-sentedtheriskthattheinfectionwastransmittedfromtheinfected cowtooneoftheother99‘susceptible’cowsintheherdduringa 12-weekperiod.The12-weekperiodwassplitinto14-dayblocks meaningtheinfected cowcouldinfectanothercowintheherd every14-days.Ifinfectiondidspreadtoanothercow,thenittoo wouldbeconsideredtobeinfectiousduringthesubsequent 14-dayblocks.Forexample,ifacowremainedsubclinicallyinfected afterCM1,andittransmittedaninfectiontoanothercowduring thefirst14-dayblock,thentherewouldbetwoinfectiouscowsat thestartofthesecond14-dayblock,andthesusceptiblepopulation wouldbe98cows.However,itshouldbenotedthatthe transmis-sionparameterwouldbeapplicabletoherdsofdifferentsizesand theresultsofthemodelcouldbeextrapolatedtoherdsofanysize.

2.2. Modelinputparameters

The model was parameterized with distributions based on existing literatureand current commercialdata wherepossible

(Table1).Allparameterinputswerespecifiedasuniform

distri-butionswiththepurposeofsimulatingawidevarietyofdifferent scenarioswithoutmakingassumptionsastowhichwasthemost likely.Thedistributionrangeswerebasedontheliterature wher-ever possible, but if only point estimates were available, then plausible ranges were added to the point estimate. The input parameterswerethesameasusedin thestudyby Downetal.

(2013),withtheadditionofsomeOFC-specificparametersbased

onthestudybyLagoetal.(2011a).

Economicparameterdistributionsincludedthecostofdrugs, labour,milkwithdrawalandlossofmilkproduction,cullingand death(Table 1).Thecost of labouris subject tolargevariation quotedintheliterature.Forthisreasonawidedistributionwas assignedtothehourlycostoflabourwiththeupperlimittaken

fromHuijpsetal.(2008).Thetotaltimetakentotreateachcase

ofCM wasassigned adistributioncentredonthefigures given

bySteeneveldetal.(2011)surroundedbyanadditionalvariation

of+/−10minutes.Thetotalcostoflabourwastheproductofthe hourlyrateandthetotaltreatmenttime.

ThelengthofmilkwithdrawalafterCMwasdefinedbya distri-butionbasedonthecommonlyusedmedicinesintheUKandthe amountofmilkbeingdiscardedeachdaywastakenfromaplausible milkyielddistribution(Table1).Thedistributiondefinedformilk pricewastakenfromDairyCo(2012a)andbasedontheaverageUK milkpriceoverthelast12months(range:lowestpriceandhighest price).ThecostofmilkproductionwasbasedonHuijpsetal.(2008)

(3)

Fig.1.Schematicrepresentationofthetreatmentmodel.Completecure=bacteriologicalandclinicalcure.Clinicalcure=non-bacteriologicalcurebutclinicalcure.No cure=nobacteriologicalorclinicalcure.Cull=culledsometimewithintheremainderofthecurrentlactation.Extendedtreatment=arepeatofthesametreatmentthatthe casereceivedinitially.RT=riskoftransmission.CM1=initialcaseofclinicalmastitis.CM2=firstclinicalflare-up.CM3=seconclinicalflare-up.

ThecalculationoftotalyieldlossfollowingacaseofCMwas based onthe herd 305day yield, the parity of theanimal and thestageoflactationinwhich theinfectionoccurred(Table1). Thedistributionsgoverningthepercentageoftotallossin305day milkyieldwerebasedonHagnestametal.(2007).Theproportion of cases occurring at each stage postpartum and the propor-tionofcasesaffectingmultiparouscowsversusprimiparouscows wasgovernedbydistributionsbasedonSteeneveldetal.(2011)

(Table1).Thecostassociatedwiththetotallossinmilkyieldwas

calculatedaccordingtothetotallossinearnings(i.e.thequantity ofmilkmultipliedbythemilkprice)minusthesavingsmadein feedcosts(i.e.thequantityofmilklossmultipliedbythecostof production).AlldistributionsareprovidedinTable1.

Thecostofcullingacowwithintheremainderofthecurrent lac-tationwastakenfromauniformdistributionbasedonHuijpsetal.

(2008)andKossaibatiandEsslemont(2000),whichincludedthe

slaughtervalueandreplacementcosts,withanappropriaterange addedtoreflectthevariabilityofthisparameter(Table1).Thecost ofthedeathofanindividualwasbasedoncurrentUKaveragesales pricesforfreshlycalvedcowsandheifers(DairyCo,2012b)which wouldberequiredtoreplacethedeadcowinadditiontothecost ofcarcassdisposal(Table1).

TheoriginalcalculationsweremadeinGreatBritishPounds(

£

) andconvertedtoUSDollars($)usingtheexchangerateof1.45$/

£

(http://www.xe.com/currencyconverter/;accessed15thFebruary

2016).

2.3. On-farmculturespecificinputparameters

The OFC-specific input parameters comprised distributions reflectingchangestothebacteriologicalcurerisk,theproportionof culture-positivecases,thetimetakentoset-upandreadtheculture platesandthecostofaplate.Thedistributionforthereductionin

bacteriologicalcureriskassociatedwiththeOFCprotocolwas uni-form(−0.22to0),meaningthemaximumreductionpossiblewas 22%,andtheminimumwas0.Themiddlevalueof11%wasthe non-significanteffectsizereportedbyLagoetal.(2011a)which wastheoverallreductioninbacteriologicalcureriskincasesof clinicalmastitistreatedwiththeOFCprotocolcomparedtocases treatedwiththeconventionalapproach.Thedistribution specify-ingthe‘herd-level’proportionofGram-positivecaseswasuniform (0.1–0.9),meaningthelowestproportionwas10%ofcaseswitha Gram-positivecultureandthehighestproportionwas90%.This dis-tributionreflectsthewidespreadofvaluesidentifiedinthestudy

byLagoetal.(2011a).Therewerenopublishedfiguresforthecost

ofthebiplateusedinthestudyorthetimetakentoset-upand evaluatethecultureresultssoplausiblerangeswereestimatedas ($1.45–2.03)and(30–60min)respectively.Thedistributionsused forallotherinputparametersarelistedinTable1.

2.4. Modelsimulation

(4)

94 P.M.Downetal./PreventiveVeterinaryMedicine145(2017)91–99

Table1

Probabilitydistributionsapplicabletobothtreatmentprotocolsandrelevantsourcesofliteratureonwhichtheyarebasedwhereapplicable.

Inputparameters Upperandlowerlimits ofuniformdistribution

Source

Probabilityofbacteriologicalcure(0.40,0.80) a,b,c,d,e,f,g,h

Probabilityofbacteriologicalcureafterextendedtx(0.30,0.90) BaseduponSteeneveldetal.(2011)

Decreaseinprobabilityofbacteriologicalcure1 BaseduponSteeneveldetal.(2011)

Parity≥2 (−0.15,−0.05)

Daysinmilk≥60days (−0.15,−0.05)

Cowissystemicallyill (−0.25,−0.15)

SCC200,000–500,000cells/mLatmostrecentrecording (−0.15,−0.05) SCC>500,000cells/mLatmostrecentrecording (−0.25,−0.15) Repeatedcase(>1stcaseincurrentlactation) (−0.25,−0.15)

Probabilityofbeingculledforbacteriologicallynoncuredcases BaseduponBaretal.(2008)

Initialcase (0,0.32)

Followingfirstflare-up(CM2) (0.04,0.36)

Probabilityofbeingculledforcompletelycuredcases BaseduponBaretal.(2008)

Initialcase (0.04,0.06)

Followingfirstflare-up(CM2) (0.10,0.20)

Followingsecondflare-up(CM3) (0.20,0.30)

Probabilityofdeathfornonclinicalcuredcases (0.04,0.06) BaseduponBaretal.(2008)

Probabilityofdrying-offquarterfornonclinicalcuredcases (0.94,0.96) BaseduponSteeneveldetal.(2011) Probabilityofbeingculledforcowswithdriedoffquarters (0.27,0.39) BaseduponSteeneveldetal.(2011) Increaseinallcullingprobabilitieswhencowissystemicallyill (0.05,0.15) BaseduponSteeneveldetal.(2011) Probabilityofclinicalflare-upforbacteriologicallynoncuredcases (0.05,0.12) m,n,o

ProbabilityoftransmissionafterCM1andCM2 (0.002,0.25) vandenBorneetal.(2010)

Proportionalyieldloss j,k

Casein1stor2ndmonthoflactation (0.07,0.09) j,k,l

Casebetweenmonths3and6 (0.03,0.08)

Caseaftermonth6 (0,0.04)

Parity≥2 (0,0.02)

305dYield(Kg) (7000,10,000) Author2

Milkwithdrawal(d) (5.00,9.00) BaseduponcommonlyusedpreparationsintheUK

Dailymilkdiscard(Kg) (5.00,50,00) Author2

Valueofdiscardedmilk($/Kg) (0.33,0.39) DairyCo(2012a)

Costofmilkproduction($/Kg) (0.043,0.145) BaseduponHuijpsetal.(2008)

TreatmentTime(hr) (0.53,0.87) BaseduponSteeneveldetal.(2011)

Costoflabour($/hr) (1.45,23.01) BaseduponHuijpsetal.(2008)

Costofdrugs($) (8.10,10.11) Baseduponestimateofcurrentretailpriceof

commonlyusedpreparationsintheUK

Costofcull($) (174,1044) BaseduponHuijpsetal.(2008),Kossaibatiand

Esslemont(2000)

Costofdeath($) (1740,2900) DairyCo(2012b)

aBaseduponMcDougall(1998),bBaseduponMcDougall(2003),cBaseduponOliveretal.(2003),dBaseduponWraight(2003),eBaseduponSérieysetal.(2005),fBased uponMcDougalletal.(2007),gBaseduponBradleyandGreen(2009),hBaseduponSoletal.(2000),jBaseduponGröhnetal.(2004),kBaseduponSchukkenetal.(2009),l BaseduponHagnestametal.(2007),mBaseduponSwinkelsetal.(2005a),nBaseduponSwinkelsetal.(2005b),oBaseduponDöpferetal.(1999).

1 Thevalueselectedfromthisdistributionwassubtractedfromthevalueselectedfromthebacteriologicalcuredistribution.

2 Authorwheretherewasnorelevantliteratureidentifiedonwhichtobasetheparameter,distributionswerebasedonbiologicallyplausiblevaluesinstead.

Table2

On-farmculturespecificmodelinputparametersandtherelevantsourcesofliteratureonwhichtheywerebasedwhereapplicable.

Inputparameters Upperandlowerlimitsofuniformdistribution Source

ProportionofGram-positivecases (0.10,0.90) BaseduponLagoetal.(2011a)

Reductioninbacteriologicalcurerisk (−0.22,0.00) BaseduponLagoetal.(2011a)

Costofplate($) (1.45,2.03) Baseduponcurrentretailprice

Culturetime(h) (0.30.1.00) Expertopinion

theOFCprotocolwasmorecost-effective.Thedistribution specify-ingtheherd-levelproportionofGram-positivecaseswouldgovern whetherthecasetreatedaccordingtotheOFCprotocolwas Gram-positive(ormixedinfection)and,therefore,treatedwithantibiotics orGram-negative(ornogrowth)and,therefore,nottreatedwith antibiotics.In this way, theimpactof the proportionof Gram-positivecasesontheoverallcost-effectivenessoftheOFCprotocol couldbeassessed.

2.5. Dataanalysis

Spearman rank correlation coefficients were calculated to exploretheunivariateassociationsbetweenmodelinput param-etersandthedifferenceincostbetweentheconventionalandOFC treatmentprotocols.Thestrengthanddirectionoftherelationships wereevaluatedusingtheSpearmanrankrho(␳)value.Theoutcome

variableofspecificinterestwasthedifferenceincostbetweenthe twotreatmentprotocols.However,additionalmodelparameters wereincludedtoprovidefurtherinsightintowherecostdifferences arose.Thesewerethecostofantimicrobialdrugs,thedifferencein timetakentotreateachcase,thedifferenceinmilkwithdrawal periodandthedifferenceintheriskoftransmission.

(5)

Fig.2. Differenceincostbetweenconventionalandon-farmcultureprotocols.Apositivedifferenceincostindicatesthattheon-farmcultureprotocolcostmorethanthe conventionalprotocol.

Fig.3.Differenceincostbetweenconventionalandon-farmcultureprotocolswhentherewasasmalldifference(SD)inbacteriologicalcurerisk.Apositivedifferencein costindicatesthattheon-farmcultureprotocolcostmorethantheconventionalprotocol.

Table3

Spearmanrankcorrelationcoefficientsforon-farmspecificmodelinputparameters andthedifferenceincostbetweentheconventionalandon-farmculturetreatment protocols.

Parameter rho

Proportionculture-positive 0.31 Differenceinbacteriologicalcurerisk −0.28

Costofplate 0.0062

Culturetime 0.02

thenassessedforthedifferentgroupsandatdifferentproportions ofGram-positivecases.

3. Results

3.1. Dataanalysis

Acrossall5000simulatedcases,theconventionalprotocolwas themostcost-effective68%ofthetime.Themediancostrelated toacasetreatedwiththeconventionalprotocolwas$529,andthe mediancostrelatedtoacasetreatedwiththeOFCprotocolwas $554.Themaximumdifferenceincostbetweenthetwoprotocols was$328withamedianof$28.

TheSpearmanrankcorrelationcoefficientsfortheOFC-specific parametersareshowninTable3.Thedifferenceincostbetween

thetwoprotocolswasmostcloselyrelatedtothedifferencein bac-teriologicalcureriskandtheproportionofGram-positivecases.As thedifferenceinbacteriologicalcureriskandproportionof Gram-positive cases increased, the difference in overall cost became higher,makingtheOFCprotocollesscost-effectivethanthe con-ventionalprotocol.Boththecostofthebiplateandthetimetaken toset-upand evaluatethebiplatehada negligiblerelationship withthecost-effectivenessoftheOFCprotocolasmeasuredbythe Spearmanrankcorrelationcoefficients(Table3).

Withrespecttothemodelinputparameterscommontoboth protocols,thosesignificantlyassociatedwiththedifferenceincost werethedifferenceinthemilkwithdrawalperiod(rho=0.75), dif-ferenceinthecostofdrugs(rho=0.61),differenceinthetimetaken totreatthecow(andculture)(rho=0.61)andthedifferenceinthe riskoftransmission(rho=0.51).

3.2. Scenarioandsensitivityanalysis

(6)

96 P.M.Downetal./PreventiveVeterinaryMedicine145(2017)91–99

Fig.4.Differenceincostbetweenconventionalandon-farmcultureprotocolswhentherewasamoderatedifference(MD)inbacteriologicalcurerisk.Apositivedifference incostindicatesthattheon-farmcultureprotocolcostmorethantheconventionalprotocol.

Fig.5.Differenceincostbetweenconventionalandon-farmcultureprotocolswhentherewasalargedifference(LD)inbacteriologicalcurerisk.Apositivedifferencein costindicatesthattheon-farmcultureprotocolcostmorethantheconventionalprotocol.

average$42moreexpensivepercasethantheconventional proto-col.However,thedifferenceincostbetweenthetreatmentgroups wassensitivetotheunderlyingbacteriologicalcureriskof Gram-positivecases.Whenclinicalmastitiswassubdividedaccordingto whetherthedifferenceinbacteriologicalcureriskwassmall(SD) medium(MD)orlarge(LD)thedifferenceinthecost-effectiveness ofthetreatmentswasasfollows.TheOFCprotocolwasmore cost-effectivethantheconventionalprotocolwhentheproportionof Gram-positivecaseswaslessthan47%intheSDgroup(Fig.3)and lessthan21%intheMDgroup(Fig.4).TheOFCprotocolwasnever morecost-effectivethantheconventionalprotocolforcasesinthe LDgroup(Fig.5).Therefore,theunderlyingbacteriologicalcurerisk wasakeyparameterdeterminingrelativecost-effectivenessofthe treatmentapproaches.

4. Discussion

Thesimulationanalysesrevealedthatboththedifferenceinthe bacteriologicalcureriskduetoadelayintreatmentandthe

pro-portionofGram-positivecasesthatoccuronadairyunitwillhave afundamentalimpactonwhetherOFCwillbecost-effective.There hasundoubtedlybeenashiftintheaetiologyofclinicalmastitis towardsenvironmentalpathogens,withcoliformsandno-growths frequentlyreportedasaccountingforapproximately50%ofall clin-icalmastitiscultureresults(BradleyandGreen,2001;Bradleyetal.,

2007;Breenetal.,2009)aswasthecaseinthestudybyLagoetal.

(2011a).Onthisbasis,itwouldbefairtoassumethatmostdairy

(7)

furtherworktoquantifyanylikelyreductioninbacteriologicalcure riskthatisassociatedwithOFCiscritical,ifthecost-effectiveness andwelfareimplicationsofOFCaretobeestablished.

OneoftheaimsstatedbytheauthorsoftheoriginalOFCstudies

(Lagoetal.,2011a,b)wastousetheirresultstoevaluatethe

over-allcost-benefitofusinganOFCsystem,buttodate,nodatahave beenpublished.Inadecisiontreeanalysisusedtoevaluatethe eco-nomicimpactofdifferentdurationsofintramammarytreatment forthefirstcase ofmildormoderateclinicalmastitisoccurring inearlylactation,itwasfoundthatOFCwasonlylikelytooffer economicbenefitsinherdsusingextended-durationtherapy with-outregardforpathogendiagnosis(Pinzón-Sánchezetal.,2011). TheresultsofthisstudyservetoillustratethatanOFCapproach forthetreatmentofclinicalmastitiswouldprobablynotbe cost-effectiveinmanycircumstances,inparticular,notthoseinwhich Gram-positivepathogensrepresentmorethan20%ofallclinical cases.SinceStreptococcusuberisandStaphylococcusaureusremain commonmastitispathogensondairyunitsinmanycountries,the cost-effectivenessofOFCshouldbecarefullyscrutinisedinthese circumstances.

WhileOFCwillreducetotalantimicrobialdrugusageonfarm, theeffectoncowhealthandwelfareandoveralldairyfarm prof-itabilityshouldalsobeconsidered.Theassertionthatthereisno importantreductioninbacteriologicalcurefromdelayedtreatment ofGram-positive pathogensis fragile andrequires substantially moreresearchwithsufficient powertodetectsmalldifferences ineffectsize.InthestudybyLagoetal.(2011a),statistical anal-ysisrevealedanon-significantdifferenceof11%inbacteriological cureriskbetweentheconventionalandOFCgroups.Inthatstudy, thesamplesizeusedmeantthat adifference≥14%wouldhave beenneededbetweengroupstodetectthedifferenceasbeing ‘sig-nificant’(Lagoetal.,2011a)andit,therefore,remainsuncertain whetherthereisatruedifferenceinbacteriologicalcurebetween groups.Thesensitivityanalysisinthecurrentstudysuggeststhat a differencein curerisk ofless than14%couldcertainly deter-minewhetherOFCiscost-effectiveand,therefore,largerstudies toascertainthistruedifferenceareneeded.

Significantdifferenceswerereportedinthepathogen-specific bacteriologicalcureriskinthestudybyLagoetal.(2011a), partic-ularlywithrespecttoStaphylococcusaureus.Whilethereasonfor thesedifferencesisunknown,itispossiblethatthereductionin bacteriologicalcureriskassociatedwithOFCisaresultofdelayed treatmentaswashypothesisedbyLagoetal.(2011a)andhasbeen reportedinapreviousstudy(HillertonandSemmens,1999).Given theimportanceofthisparameter,futureresearchshouldinclude pathogen-specific differences in bacteriological cure risk, when treatmentisdelayedbyusingOFC.

Inthecurrentstudy,theoverallproportionofGram-positive cases was also shown to be related to the likelihood of cost-effectivenessofanOFCtreatmentprogramme.Theproportionof Gram-positivecaseswasshowntobehighlyvariableinthestudy

byLago etal.(2011a) inwhich theproportionofquartercases

receivingintramammaryantibiotic treatmentasa consequence ofassignmenttotheOFCprotocolrangedfrom31%–89%inthe eightstudyherds.Inthecurrentstudy,theoverallproportionof Gram-positivecaseshadtobelessthan12%(dependingon bac-teriologicalcurerisk)forOFCtobemorecost-effectivethanthe conventionalprotocol.Bythismeasure,OFCwouldnothavebeen cost-effectiveinanyoftheherdsinthestudybyLagoetal.(2011a). However,whencasesweregroupedaccordingtothedifferencein bacteriologicalcurerisk,OFCwouldbecost-effectivewhenthe pro-portionofGram-positivecaseswaslessthan47%intheSDgroup andlessthan21%intheMDgroup.TheOFCapproachwould, there-fore,bemostsuitableforherdsinwhichGram-negativepathogens areresponsibleformostclinicalmastitisandwherethetreatment ofcowsusinganOFCapproachresultsinaminimalreductionin

thebacteriologicalcurerisk.Inpractice,itispossibletoassessthe proportionofGram-positivecasesonaunit,andthiswillinform decisionmakingonthelikelycost-benefitofOFC.

Theremaybeabalancetobestruckbetweenreducing antimi-crobial usage and possible deleterious effects in terms of cow welfareandfarmfinances;wouldtheextracostincurredby adopt-inganOFCapproachbeconsideredapriceworthpayingifitresults inareductioninantibioticdrugusageondairyfarmsby25%,aswas estimatedbyLagoetal.(2011a)?If,forsocietalreasons,thiswas consideredtobeapriceworthpaying,thereisalsoanissueofwho shouldbearthecost.Whilstdifficult,itisperhapstimesuchdebates becametransparentgiventheincreasingpressureonantimicrobial drugusageandthepotentialrisksposedbyantimicrobial resis-tantbacteria.Intheabsenceoflegaljurisdiction,itisincumbent onthose advisingonanimal health and welfare toensurethat theadoptionofnewtechnologies,suchasOFC,areundertakenin lightofcomprehensive,transparentwelfareandcost-effectiveness assessments.

Whilsttheoveralllikelihoodofcost-effectivenesswasaffected mostlybytheproportionofGram-positivecasesandthedifference inbacteriologicalcurerisk,theparameterswithinthemodelthat hadthelargestimpactonthedifferenceincostwerethe differ-enceinmilk-withdrawalperiod,thedifferenceinthecostofdrugs, thedifferenceincultureandtreatmenttimeandthedifferencein riskoftransmission.Clearly,OFCwouldbeexpectedtoreducethe amountofmilkwithdrawnfromsaleandtheamountofmoney spentondrugsbecauseaproportionofthecowswouldnotreceive anyantimicrobialtreatmentandwould thereforenot incurany statutorymilkwithholduponresolutionofclinicalsigns.Thisis inagreementwithLagoetal.(2011a)thatreportedareductionin milkwithdrawalperiod(5.2daysv5.9days)andquantityof antimi-crobialdrugusage(51%ofOFCcasestreatedv100%ofconventional casestreated)associatedwithOFC.Theincreaseinlabourrequired toacquiremilksamplesfromclinicalmastitiscasesinanaseptic mannerandplateoutforcultureisperhapshardertoassessandis likelytorepresentacostnotonlyintermsofthetimetaken,but alsotheopportunitycostincurredasaresultoftheherdsmanbeing unabletoperformotherdutiesasaresult.Thedistributionusedin thisstudyof30–60minis,therefore,likelytobearealisticestimate formostcircumstances.Thelargeimpactthattransmissioncould haveonthecostofacaseofclinicalmastitishasbeenreported pre-viously(Downetal.,2013)anditisnotsurprisingthereforethat itwascloselyrelatedtothedifferenceincostbetweenthe con-ventionalandOFCapproachesalso.Whiletheriskwouldclearlybe influencedbyherdmanagementandpathogen-specificfactors;it couldalsobeaffectedbyanydelayintreatmentanddifferences in bacteriological curerisk associated withOFC resulting inan increasedriskoftransmission.Thisagainwouldneedtobeassessed attheherdlevel.Itisworthnotingthatmilkpricehadverylittle impactonthedifferenceincostbetweenthetwoprotocols.Thisis unsurprisinggiventhepoorcorrelationbetweenmilkpriceandthe costofacaseofCMthathasbeenreportedpreviously(Downetal., 2013)andservestodemonstratethatwhilstimportantinterms ofoverallfarmprofitability(SmithandThanassoulis,2015),milk priceplaysaminorroleintermsofthecostofCM.

(8)

98 P.M.Downetal./PreventiveVeterinaryMedicine145(2017)91–99

arbitrarythresholdsof‘significance’(Gurrinetal.,2000;Greenland

andPoole,2013)meaningtheclinicianisfreetomaketheirown

judgementas towhat is clinically ‘significant’ accordingtothe degreeofuncertaintytheyarecomfortablewith.Inthisstudy,the PSAallowedanevaluationoftheparameterslikelytobeimportant indeterminingthecost-effectivenessoftheOFCapproachandhas highlightedthatmoreresearchisneededinthisfieldbeforethe techniquecanberecommendedonawidespreadbasis.

5. Conclusions

Theresultsofthisstudyindicatethattheproportionof Gram-positive cases and the difference in bacteriological cure risk betweenthetwotreatmentapproacheshasthegreatestimpacton theprobabilitythatanOFCapproachwouldbemorecost-effective thanaconventionalapproachforthetreatmentofclinical masti-tis.TheOFCapproachappearstobesuitable forherdsin which Gram-negativepathogensareresponsibleformostclinical masti-tisandwherethetreatmentofcowsaccordingtotheresultsofan OFCapproachresultsinminimalreductionsinthebacteriological curerisk.TheseresultssuggestthatOFCwillprobablynotbe cost-effectiveformanyherdsandthatOFCshould,therefore,onlybe adoptedaftercarefulconsiderationofthepredominantpathogens presentineachherdandanhonestdiscussionabouttheuncertainty surroundingitsoverallcost-effectiveness.

Conflictofintereststatement

Theauthorshavenoconflictsofinterest

Acknowledgements

Thisworkwassupportedby theBiotechnologyand Biologi-calSciencesResearchCouncil[grantnumberBB/I015493/1];and AHDBDairy. The study sponsors werenot directlyinvolved in studydesign,analysisorinterpretationofdata,inthewritingof themanuscript,orinthedecisiontosubmitthemanuscript for publication.

References

Ades,A.E.,Sculpher,M.,Sutton,A.,Abrams,K.,Cooper,N.,Welton,N.,Lu,G.,2006. Bayesianmethodsforevidencesynthesisincost-effectivenessanalysis. Pharmacoeconomics24,1–19.

BaboMartins,S.,Rushton,J.,2014.Cost-effectivenessanalysis:addingvalueto assessmentofanimalhealthwelfareandproduction.Rev.Sci.Tech.33, 681–689.

Bar,D.,Gröhn,Y.T.,Bennett,G.,González,R.N.,Hertl,J.A.,Schulte,H.F.,Tauer,L.W., Welcome,F.L.,Schukken,Y.H.,2008.Effectsofrepeatedepisodesofgeneric clinicalmastitisonmortalityandcullingindairycows.J.DairySci.91, 2196–2204.

Boshuizen,H.C.,vanBaal,P.H.M.,2009.Probabilisticsensitivityanalysis:bea Bayesian.ValueHealth12,1210–1214.

Bradley,A.J.,Green,M.J.,2001.AetiologyofclinicalmastitisinsixSomersetdairy herds.Vet.Rec.148,683–686.

Bradley,A.J.,Green,M.J.,2009.Factorsaffectingcurewhentreatingbovineclinical mastitiswithcephalosporin-basedintramammarypreparations.J.DairySci. 92,1941–1953.

Bradley,A.J.,Leach,K.A.,Breen,J.E.,Green,L.E.,Green,M.J.,2007.Surveyofthe incidenceandaetiologyofmastitisondairyfarmsinEnglandandWales.Vet. Rec.160,253–258.

Breen,J.E.,Green,M.J.,Bradley,A.J.,2009.Quarterandcowriskfactorsassociated withtheoccurrenceofclinicalmastitisindairycowsintheUnitedKingdom.J. DairySci.92,2551–2561.

Briggs,A.H.,Goeree,R.,Blackhouse,G.,O’Brien,B.J.,2002.Probabilisticanalysisof cost-effectivenessmodels:choosingbetweentreatmentstrategiesfor gastroesophagealrefluxdisease.Med.Decis.Making22,290–308.

Brown,J.,Welton,N.J.,Bankhead,C.,Richards,S.H.,Roberts,L.,Tydeman,C.,Peters, T.J.,2006.ABayesianapproachtoanalysingthecost-effectivenessoftwo primarycareinterventionsaimedatimprovingattendanceforbreast screening.HealthEcon.15,435–445.

Buxton,M.J.,Drummond,M.F.,VanHout,B.A.,Prince,R.L.,Sheldon,T.A.,Szucs,T., Vray,M.,1997.Modellingineconomicevaluation:anunavoidablefactoflife. HealthEcon.6,217–227.

Claxton,K.,Sculpher,M.,McCabe,C.,Briggs,A.,Akehurst,R.,Buxton,M.,Brazier,J., O’Hagan,T.,2005.ProbabilisticsensitivityanalysisforNICEtechnology assessment:notanoptionalextra.HealthEcon.14,339–347.

Döpfer,D.,Barkema,H.W.W.,Lam,T.J.G.M.J.,Schukken,Y.H.H.,Gaastra,W.,1999. RecurrentclinicalmastitiscausedbyEscherichiacoliindairycows.J.DairySci. 82,80–85.

DairyCo,2012a.GBFarmgateMilkPrices[WWWDocument].URLhttp://www. dairyco.org.uk/market-information/milk-prices-contracts/farmgate-prices/ uk,-gb-and-ni-farmgate-prices/.

DairyCo,2012.http://www.dairyco.org.uk/market-information/farm-expenses/ cow-heifer-prices/gb-cow-heifer-prices/[WWWDocument].URLhttp://www. dairyco.org.uk/library/market-information/datum/hygienic-quality.aspx. Down,P.M.,Green,M.J.,Hudson,C.D.,2013.Rateoftransmission:amajor

determinantofthecostofclinicalmastitis.J.DairySci.96,6301–6314. González,S.M.,Steiner,A.,Gassner,B.,Regula,G.,2010.AntimicrobialuseinSwiss

dairyfarms:quantificationandevaluationofdataquality.Prev.Vet.Med.95, 50–63.

Gröhn,Y.T.,Wilson,D.J.,González,R.N.,Hertl,J.A.,Schulte,H.,Bennett,G., Schukken,Y.H.,2004.Effectofpathogen-specificclinicalmastitisonmilkyield indairycows.J.DairySci.87,3358–3374.

Green,M.J.,Hudson,C.D.,Breen,J.E.,Bradley,A.J.,2009.Thetruecostsofmastitis. In:Br.Mastit.Conf.2009,StoneleighPark,Warwickshire,UK,14thOct.2009. Greenland,S.,Poole,C.,2013.Livingwithpvalues:resurrectingaBayesian

perspectiveonfrequentiststatistics.Epidemiology24,62–68. Gurrin,L.C.,Kurinczuk,J.J.,Burton,P.R.,2000.Bayesianstatisticsinmedical

research:anintuitivealternativetoconventionaldataanalysis.J.Eval.Clin. Pract.6,193–204.

Hagnestam,C.,Emanuelson,U.,Berglund,B.,2007.Yieldlossesassociatedwith clinicalmastitisoccurringindifferentweeksoflactation.J.DairySci.90, 2260–2270.

Hillerton,J.E.,Semmens,J.E.,1999.Comparisonoftreatmentofmastitisby oxytocinorantibioticsfollowingdetectionaccordingtochangesinmilk electricalconductivitypriortovisiblesigns.J.DairySci.82,93–98. Hudson,C.D.,Huxley,J.N.,Green,M.J.,2014.Usingsimulationtointerpreta

discretetimesurvivalmodelinacomplexbiologicalsystem:fertilityand lamenessindairycows.PLoSOne9,e103426.

Hudson,C.D.,Bradley,A.J.,Breen,J.E.,Green,M.J.,2015.Dairyherdmastitisand reproduction:usingsimulationtoaidinterpretationofresultsfromdiscrete timesurvivalanalysis.Vet.J.204,47–53.

Huijps,K.,Lam,T.J.,Hogeveen,H.,2008.Costsofmastitis:factsandperception.J. DairyRes.75,113–120.

Kossaibati,M.A.,Esslemont,R.J.,2000.ThecostsofclinicalmastitisinUKdairy herds.CattlePract.8,323–327.

Lago,A.,Godden,S.M.,Bey,R.,Ruegg,P.L.,Leslie,K.,2011a.Theselectivetreatment ofclinicalmastitisbasedonon-farmcultureresults:I.Effectsonantibioticuse milkwithholdingtime,andshort-termclinicalandbacteriologicaloutcomes.J. DairySci.94,4441–4456.

Lago,A.,Godden,S.M.,Bey,R.,Ruegg,P.L.,Leslie,K.,2011b.Theselectivetreatment ofclinicalmastitisbasedonon-farmcultureresults:II.Effectsonlactation performanceincludingclinicalmastitisrecurrence,somaticcellcount,milk production,andcowsurvival.J.DairySci.94,4457–4467.

McDougall,S.,Arthur,D.G.,Bryan,M.A.,Vermunt,J.J.,Weir,A.M.,2007.Clinicaland bacteriologicalresponsetotreatmentofclinicalmastitiswithoneofthree intramammaryantibiotics.N.Z.Vet.J.55,10.

McDougall,S.,1998.Efficacyoftwoantibiotictreatmentsincuringclinicaland subclinicalmastitisinlactatingdairycows.N.Z.Vet.J.46,7.

McDougall,S.,2003.Intramammarytreatmentofclinicalmastitisofdairycows withacombinationoflincomycinandneomycin,orpenicillinand dihydrostreptomycin.N.Z.Vet.J.51,6.

O’Hagan,A.,2003.APrimeronBayesianStatisticsinHealthEconomicsand OutcomesResearch.MedtapIntInc,London.

O’Neill,J.,2015.AntimicrobialsinAgricultureandtheEnvironment:Reducing UnnecessaryUseandWaste.

Oliver,S.P.,Almeida,R.A.,Gillespie,B.E.,Ivey,S.J.,Moorehead,H.,Lunn,P.,Dowlen, H.H.,Johnson,D.L.,Lamar,K.C.,2003.Efficacyofextendedpirlimycintherapy fortreatmentofexperimentallyinducedStreptococcusuberisintramammary infectionsinlactatingdairycattle.Vet.Ther.4,299–308.

Pinzón-Sánchez,C.,Cabrera,V.E.,Ruegg,P.L.,2011.Decisiontreeanalysisof treatmentstrategiesformildandmoderatecasesofclinicalmastitisoccurring inearlylactation.J.DairySci.94,1873–1892.

Pol,M.,Ruegg,P.L.,2007.Treatmentpracticesandquantificationofantimicrobial drugusageinconventionalandorganicdairyfarmsinWisconsin.J.DairySci. 90,249–261.

Sérieys,F.,Raguet,Y.,Goby,L.,Schmidt,H.,Friton,G.,2005.Comparativeefficacyof localandsystemicantibiotictreatmentinlactatingcowswithclinicalmastitis. J.DairySci.88,93–99.

Schukken,Y.H.,Hertl,J.,Bar,D.,Bennett,G.J.,González,R.N.,Rauch,B.J., Santisteban,C.,Schulte,H.F.,Tauer,L.,Welcome,F.L.,Gröhn,Y.T.,2009.Effects ofrepeatedgram-positiveandgram-negativeclinicalmastitisepisodeson milkyieldlossinHolsteindairycows.J.DairySci.92,3091–3105. Smith,H.,Thanassoulis,J.,2015.Pricesprofits,andpass-throughofcostsalonga

(9)

Sol,J.,Sampimon,O.C.,Barkema,H.W.,Schukken,Y.H.,2000.Factorsassociated withcureaftertherapyofclinicalmastitiscausedbyStaphylococcusaureus.J. DairySci.83,278–284.

Steeneveld,W.,vanWerven,T.,Barkema,H.W.,Hogeveen,H.,2011.Cow-specific treatmentofclinicalmastitis:aneconomicapproach.J.DairySci.94,174–188. Swinkels,J.M.,Hogeveen,H.,Zadoks,R.N.,2005a.Apartialbudgetmodelto

estimateeconomicbenefitsoflactationaltreatmentofsubclinical Staphylococcusaureusmastitis.J.DairySci.88,4273–4287.

Swinkels,J.M.,Rooijendijk,J.G.,Zadoks,R.N.,Hogeveen,H.,2005b.Useofpartial budgetingtodeterminetheeconomicbenefitsofantibiotictreatmentof chronicsubclinicalmastitiscausedbyStreptococcusuberisorStreptococcus dysgalactiae.J.DairyRes.72,75–85.

Thomas,A.,O’Hara,B.,Ligges,U.,Sturtz,S.,2006.MakingBUGSopen.RNews6, 12–17.

Thomson,K.,Rantala,M.,Hautala,M.,Pyörälä,S.,Kaartinen,L.,2008. Cross-sectionalprospectivesurveytostudyindication-basedusageof antimicrobialsinanimals:resultsofuseincattle.BMCVet.Res.4,15. vandenBorne,B.H.P.,Halasa,T.,vanSchaik,G.,Hogeveen,H.,Nielen,M.,2010.

Bioeconomicmodelingoflactationalantimicrobialtreatmentofnewbovine subclinicalintramammaryinfectionscausedbycontagiouspathogens.J.Dairy Sci.93,4034–4044.

References

Related documents

This excavation demonstrated that an extensive area of Mesolithic occupation deposits was preserved at Bolsay Farm and consequently the 1992 excava-.. tion aimed to acquire data for

des Caspase-Inhibitors zVAD verzögerten den durch Che und Sts induzierten apoptotischen Zelltod: 20-40% mehr Zellen überlebten innerhalb der ersten sechs Stunden der Inkubation,

Background: Councils of Chiropractic Education (CCE) indirectly influence patient care and safety through their role of ensuring the standards of training delivered by

Thus, this study is designed to highlight the status of uncontrolled hypertension in patients with type 2 diabetes and determine the associated factors, which may affect the

Considering only women who received abortion care from public facility where there is shortage contraception supplies in study conducted in two of the region and failing to

It was decided that with the presence of such significant red flag signs that she should undergo advanced imaging, in this case an MRI, that revealed an underlying malignancy, which

Via the unique personal identifier, used in all official records in Denmark, 11 we linked the GP-reported treatment data to the patients’ records in the Aarhus

In contrast, Rao et al found peripapillary (AUC =0.83) and inside disc VD (AUC =0.73) to be more Figure 3 Correlation between inner retinal thickness and parafoveal vessel