exhibit 94 · glyphosate acid is a colorless, odorless, crystalline solid. glyphosate is the term...

EXHIBIT 94

Case 3:16-md-02741-VC Document 654-17 Filed 10/28/17 of 97

ExpertReportChristopherJ.Portier,Ph.D.

Charge

Glyphosateacidisacolorless,odorless,crystallinesolid.Glyphosateisthetermusedtodescribethesaltthatisformulatedbycombiningthedeprotonatedglyphosateacidandacation(isopropylamine,ammonium,orsodium).Thisexpertreportisintendedtoreviewtheavailablescientificevidencerelatingtothepotentialofglyphosateandglyphosate-basedformulations(GBFs),includingRoundup®,tocauseNon-Hodgkin’sLymphoma(NHL)inhumans.

Qualifications

Ireceivedanundergraduatedegreeinmathematicsin1977fromNichollsStateUniversityandaMaster’sdegreeandPh.D.inbiostatisticsfromtheUniversityofNorthCarolinaSchoolofPublicHealthin1979and1981respectively.MyPh.D.thesisaddressedtheoptimalwaytodesignatwo-yearrodentcarcinogenicitystudytoassesstheabilityofachemicaltocausecancer[1,2];theoptimaldosingpatternfrommythesisisstillusedbymostresearchers.MyfirstemploymentfollowingmydoctoraldegreewasajointappointmentattheNationalInstituteofEnvironmentalHealthSciences(NIEHS)andtheNationalToxicologyProgram(NTP)toconductresearchonthedesignandanalysisofexperimentsgenerallyemployedintoxicology.After5yearswithNIEHS/NTP,IdevelopedmyownresearchgroupwhicheventuallybecametheLaboratoryofQuantitativeandComputationalBiologyandthentheLaboratoryofComputationalBiologyandRiskAssessment(LCBRA).OnehighlightduringthisperiodwasthedevelopmentofthePoly-3Testforsurvivaladjustmentofdatafromtwo-yearcarcinogenicitystudiesinrodents[3,4];thistestisusedasthemainmethodofanalysisofthesestudiesbytheNTPandmanyothers.WealsodidacompleteanalysisofthehistoricalcontrolsanimalsfromtheNTPstudies[5,6].TheLCBRAfocusedontheapplicationofcomputationaltoolstoidentifychemicalsthataretoxictohumans,todeveloptoolsforunderstandingthemechanismsunderlyingthosetoxicitiesandtoquantifytheriskstohumansassociatedwiththesetoxicities.ThemaintoxicologicalfocusoftheLCBRAwascancerandmylaboratorydevelopedmanymethodsforapplyingmultistagemodelstoanimalcancerdataandimplementedtheuseofthesemodelsinseveralexperimentalsettings[7-19].InmylastfewyearsattheNIEHS/NTP,myresearchfocusexpandedtothedevelopmentoftoolsforevaluatingtheresponseofcomplexexperimentalandhumansystemstochemicals[20-24]andthenameofthelaboratoryshiftedtoEnvironmentalSystemsBiology.

Overmy32yearswiththeNIEHS/NTP,Iwasinvolvedinnumerousnationalpriorityissuesthatwentbeyondmyindividualresearchactivities.AfterCongressaskedNIEHStoworkwiththeVietnamesegovernmenttoaddressthehazardsassociatedwithAgentOrangeuseduringtheVietnameseWar,Iwasgiventheresponsibilityofworkingwith


2

mycounterpartsinVietnamtobuildaresearchprograminthisarea[25].CongressalsotaskedNIEHSwithdevelopingaresearchprogram(EMF-RAPID)toaddressconcernsabouttheriskstohumansfromexposuretopowerlinesandtoreportbacktoCongressonwhatwefound.IwasinchargeofevaluatingallresearchdevelopedunderthisprogramandwasresponsibleforthefinalrecommendationstoCongressonthisissue[26-28].

WhileattheNIEHS/NTP,Ialsohadadministrativepositionsthatrelatetomyqualifications.From2000to2006IwastheDirectoroftheEnvironmentalToxicologyProgram(ETP)atNIEHS.TheETPincludedallofthetoxicologyresearchlaboratorieswithintheNIEHSIntramuralResearchProgram.ItwasmyresponsibilitytoensuretheresearchbeingdonewaspertinenttothemissionoftheNIEHS,addressinghighpriorityconcernsabouttoxicsubstancesandhumanhealthandthattheNIEHShadadequateresourcestocompletethisresearch.

DuringthistimeIwasalsoAssociateDirectoroftheNTP,apositioninwhichIwasthescientificandadministrativedirectoroftheNTP(TheDirectoroftheNTPwasalsotheNIEHSDirectorandgavemecompleteautonomyinthemanagementandscienceoftheNTP).ThesetwopositionswerehistoricallyalwayscombinedattheNIEHSandtheNTPsothatonepersonwasinchargeofalltoxicologicalresearchattheNIEHS/NTP.TheNTPistheworld’slargesttoxicologyprogram,routinelyhaving15to25activetwo-yearcarcinogenicitystudies,numerousgenetictoxicologystudiesandmanyothertoxicologicalstudiesbeingconductedatanygiventime.TheNTPtwo-yearcarcinogenicitystudiesandtheirtechnicalreportsarealsoconsideredthe“goldstandard”ofcancerstudiesduetotheirextremehighquality,theirtremendousutilityinevaluatinghumanhealthhazardsandtherigorandtransparencytheybringtotheevaluationofthedata.AlldatafromNTPtwo-yearcancerstudiesarepubliclyavailableincludingdataonindividualanimalsandimagesfromthepathologyreviewofeachanimal.TheNTPisalsohometotheReportonCarcinogens,theUSDepartmentofHealthandHumanServicesofficiallistofwhatisknownorreasonablyanticipatedtobecarcinogenictohumans.ItwasmyresponsibilitytodecidewhatitemseventuallywentontothislistwhileIwasAssociateDirectoroftheNTP.In2006,IbecameanAssociateDirectoroftheNIEHS,asenioradvisortothedirectorandthedirectoroftheOfficeofRiskAssessmentResearch(ORAR).ORARfocusedonstimulatingnewresearchareasontheevaluationofhealthrisksfromtheenvironmentandaddressedmajorriskassessmentissuesonbehalfoftheNIEHS/NTP.Forexample,inthiscapacity,Ileadamultiagencyefforttounderstandthehealthriskstohumansfromclimatechangeandtodeveloparesearchprograminthisarea[29].

IlefttheNIEHS/NTPin2010tobecometheDirectoroftheNationalCenterforEnvironmentalHealth(NCEH)attheCentersforDiseaseControlandPreventionandsimultaneouslyDirectoroftheAgencyforToxicSubstancesandDiseaseRegistry(ATSDR).NCEHdoesresearchandsupportsactivitiesaimedatreducingtheimpactofenvironmentalhazardsonpublichealth.Onewell-respectedresearcheffortoftheNCEHistheNationalBiomonitoringProgram.Thisprogramtestsforthepresenceofhundredsofchemicalsinhumanbloodandurineinanationalsampleofpeopleinthe


3

UnitedStates.ATSDRadvicestheEnvironmentalProtectionAgency(EPA)andcommunitiesonthepotentialhealthimpactsfromtoxicwastedumpsites(superfundsites).ATSDRisrequiredbylawtoproduceToxProfiles.Thesearecomprehensivereviewsofthescientificliteratureforspecificchemicalsgenerallyfoundatsuperfundsites.Theyalsoprovideanassessmentofthesafetyofthesechemicals.AspartofmyactivitiesatATSDR,IbeganamodernizationoftheToxProfilestousesystematicreviewmethodsintheirassessments;thiseffortwaslinkedtoasimilareffortthatIhadhelpedtoimplementattheNIEHS/NTP.

Asidefrommyofficialdutiesinmyvariousfederaljobs,Ialsoservedonnumerousnationalandinternationalscienceadvisorypanels.Mostnotable,formyqualificationsforthisstatement,aremyservingasChairfrom2005to2010oftheSubcommitteeonToxicsandRiskofthePresident’sNationalScienceandTechnologyCouncil,memberandchairofEPA’SScienceAdvisoryPanelfrom1998to2003(focusedspecificallyonadvisingtheirpesticidesprogram)andchairoftheInternationalAgencyforResearchonCancer(IARC)advisorygroupthatupdatedandimproveditsrulesforreviewingscientificdatatoensurethatconclusionsonthecarcinogenicityofhumanexposuresarethebestpossible(Preamble)[30].Aspartofmyworkonscienceadvisorypanels,IhaveservedonEPA’sScienceAdvisoryBoard,asanadvisortotheAustralianHealthCouncilonriskassessmentmethods,asanadvisortotheKoreanFoodandDrugAdministrationontoxicologicalmethods,andservedonseveralWorldHealthOrganization(WHO)InternationalProgramonChemicalSafetyscientificpanelsdealingwithriskassessment.BesidestheguidelinesforevaluatingcancerhazardsusedbytheIARC,IhaveeitherchairedorservedasamemberofscientificpanelsdevelopingguidancedocumentsforotherorganizationsincludingtheEPA.

Ihavereceivednumerousawards,mostnotablytheOutstandingPractitionerAwardfromtheInternationalSocietyforRiskAnalysisandthePaperoftheYearAward(twice)fromtheSocietyofToxicologyRiskAssessmentSpecialtySection.IamafellowoftheAmericanStatisticalAssociation,theInternationalStatisticalInstitute,theWorldInnovationFoundationandtheRamazinniInstitute.Ihavepublishedover250peer-reviewedscientificpapers,bookchaptersandtechnicaldocumentsontopicsintoxicologyandriskassessment.

Finally,Ihaveservedonnumerousnationalandinternationalcommitteestaskedwithevaluatingtheriskand/orhazardofspecificenvironmentalchemicals,includingglyphosate.Forexample,IhavecontributedtoriskassessmentsforEPA,theFoodandDrugAdministration,theCentersforDiseaseControlandPrevention,theNationalInstitutesofHealth,theWHOandIARC.

RelianceList

Duringthecourseofmypreparationforthisreport,Ihavereviewedthefollowingmaterials:

a. AllepidemiologicaldatarelatingtotheabilityofglyphosateformulationstocauseNHLinhumans.


4

b. ScientificpapersonthecellularoriginsofNHLc. Peer-reviewedscientificdatarelatingtothecarcinogenicity,genotoxicity

andoxidativestresscausedbyglyphosated. Technicalreportsrelatingtothecarcinogenicityofglyphosateprovidedby

thedefendanttothelawyersfortheplaintiffe. TheUSEPA,theEuropeanFoodSafetyAuthority(EFSA),theGerman

FederalInstituteforRiskAssessment,theEuropeanChemicalAgency,theIARCandtheWHO/FoodandAgricultureOrganizationJointMeetingonPesticideResiduesreviewsofthescientificliteraturerelatingtothepotentialforglyphosatetocausecancer.

f. TechnicaldocumentsavailablefromEFSAregardinganimalcarcinogenicitydataonglyphosatepreparedbyorganizationsotherthanthedefendant

g. Variousotherdocumentsproducedinthelitigation

Acompletelistofmyreliancematerialsisattheendofthisreport.

MethodologyforCausalityEvaluation

Theevaluationofwhetherglyphosateand/orGBFscancauseNHLinhumansrequiresthereviewandsynthesisofscientificevidencefromstudiesofhumanpopulations(epidemiology),animalcancerstudies,andstudiesinvestigatingthemechanismsthroughwhichchemicalscausecancer.Manydifferentapproaches[31,32]areusedtosynthesizethesethreeareasofsciencetoanswerthequestion“Doesthischemicalcausecancerinhumans?”Inanyofthesethreescienceareas,thequalityoftheindividualstudieshastobeassessedandsummarizedtomakecertainthestudiesincludedintheoverallassessmentaredoneappropriately.Oncethequalityoftheindividualstudieshasbeenassessed,ajudgmentneedstobemadeconcerningthedegreetowhichthestudiessupportafindingofcancerinhumans.Todothis,theEPA,IARC,theEuropeanChemicalAgency(EChA),theUSReportonCarcinogens,andmanyothersuseguidelines[30,33-35]thatrelyuponaspectsofthecriteriaforcausalitydevelopedbyHill(1965)[36].

Hilllistednine(9)aspectsofepidemiologicalstudiesandtherelatedsciencethatoneshouldconsiderinassessingcausality.Thepresenceorabsenceofanyoftheseaspectsisneithersufficientnornecessaryfordrawinginferencesofcausality.Instead,thenineaspectsserveasmeanstoanswerthequestionofwhetherotherexplanationsaremorecrediblethanacausalinference.AsnotedbyHill:

“Noneofmynineviewpointscanbringindisputableevidencefororagainstthecause-and-effecthypothesisandnonecanberequiredasasinequanon.Whattheycando,withgreaterorlessstrength,isto

helpustomakeupourmindsonthefundamentalquestion—isthereanyotherwayofexplainingthesetoffactsbeforeus,isthereanyotheranswerequally,ormore,likelythancauseandeffect?”

ThenineaspectscitedbyHillincludeconsistencyoftheobservedassociation,strength


5

oftheobservedassociation,biologicalplausibility,biologicalgradient,temporalrelationshipoftheobservedassociation,specificityoftheobservedassociation,coherence,evidencefromhumanexperimentationandanalogy.Thesearebrieflydescribedbelow.

Aninferenceofcausalityisstrengthenedwhenseveralofthestudiesshowaconsistentpositiveassociationbetweencancerandtheexposure.Thisaddressesthekeyissueofreplicationofstudieswhichiscriticalinmostscientificdebates.Ifstudiesarediscordant,differencesinstudyquality,potentialconfounding,potentialbiasandstatisticalpowerareconsideredtobetterunderstandthatdiscordance.

Aninferenceofcausalityisstrengthenedwhenthestrengthoftheobservedassociationinseveralstudiesarelargeandprecise.Theselarge,preciseassociationslessenthepossibilitythattheobservedassociationsareduetochanceorbias.Asmallincreaseinriskofgettingcancerdoesnotprecludeacausalinferencesinceissuessuchaspotencyandexposurelevelmayreducetheabilityofastudytoidentifylargerrisks.Meta-analysesprovideanobjectiveevaluationofthestrengthoftheobservedassociationacrossseveralstudieswithmodestriskstohelpclarifystrengthoftheobservedassociations.

Aninferenceofcausalityisstrengthenedwhenthereisdatasupportingbiologicalplausibilitydemonstratedthroughexperimentalevidence.Animalcarcinogenicitystudies,inwhichtumorincidenceisevaluatedinexperimentalanimalsexposedtopureglyphosate,playamajorroleinestablishingbiologicalplausibility.Therearenumeroustypesofmechanismsthatcanleadtocancer[37],mostofwhichcanbedemonstratedthroughexperimentalstudiesinanimals,humancells,animalcells,and/orotherexperimentalsystems.Occasionally,occupational,accidentalorunintendedexposurestohumansallowresearcherstoevaluatemechanismsusingdirecthumanevidence.

Aninferenceofcausalityisstrengthenedwhenthereisabiologicalgradientshowingareasonablepatternofchangingriskwithchangesinexposure(e.g.riskincreaseswithincreasingexposureorwithlongerexposure).Inmanyepidemiologicalstudies,thisaspectcannotbeexaminedduetolimitationsinthestudydesignorduetoalackofclarityinthepresentationoftheresults.Whenastudydoesaddressanexposure-responserelationship,failuretofindarelationshipcanbeduetoasmallrangeofexposures,insufficientsamplesizeorachangingexposuremagnitudeovertimethathasnotbeenaccountedfor.

Aninferenceofcausalityisstrengthenedwhenthereisatemporalrelationshipinwhichtheexposurecomesbeforethecancer.Thisaspectisnecessarytoshowcausality;ifitisnotpresent,acausalinferenceisnotplausible.Becausethelatencyperiodforcancerscanbelong(years),evaluationofstudiesshouldconsiderwhethertheexposureoccurredsufficientlylongagotobeassociatedwithcancerdevelopment.

Aninferenceofcausalityisstrengthenedwhentheexposureisspecificforagivencancer.Thiswouldmeanthatthediseaseendpointbeingstudiedisonlyduetothecausebeingassessed.Thisissueisseldomapplicableand,sinceNHLhasothercauses,specificityisnotapplicabletothedeterminationofcausalityforglyphosate.


6

Aninferenceofcausalityisstrengthenedwhenotherlinesofexperimentalevidencearecoherentwithacausalinterpretationoftheassociationseenintheepidemiologicalevidence.Toevaluatecoherence,informationfromanimalcarcinogenicitystudies,mechanisticinvestigationsandinformationonthemetabolismofthechemicalbeingstudiedwouldbeconsidered.

Aninferenceofcausalityisstrengthenedwhenthereisexperimentalevidenceinhumanssupportingacausalinterpretation.Seldomisthistypeofinformationavailablewhenaddressingthetoxicityofchemicals.However,experimentsinwhichanindividualreducesorlimitsexposuresandtheriskofcancerisreducedwouldcarryconsiderableweightintheevaluation(e.g.studiesevaluatingthecancerrisksofpeoplewhostopcigarettesmokingcomparedwithcontinuingsmokinghavedemonstratedreducedlungcancerrisks).Nosuchdataareavailableforglyphosate.

Finally,aninferenceofcausalityisstrengthenedwhenthereareotherchemicalagentswithanalogousstructuresshowingsimilareffectsinhumansand/oranimalsand/orshowingsimilarbiologicalimpactsinmechanisticstudies.Nosuchdataareavailableforglyphosate.

ThemostlogicalapproachtodevelopinganinferenceofcausalityistostepthrougheachoftheaspectsofcausalitydevelopedbyHill(1965)[36]andapplythemtotheavailabledataforglyphosateandforglyphosateformulations.Thisisdoneinthesectionsthatfollow.ConsistencyoftheAssociationsseeninHumanEpidemiologicalStudies

RelevantEpidemiologyStudies

Intheirmeta-analysis,ChangandDelzell(2016)[38]performedasystematicliteraturesearchofallscientificliteratureuptoJune,2015,toidentifyallepidemiologicalstudiesthatwerepertinenttoevaluatinganassociationbetweenglyphosateandNHL.Theyidentified12relevantepidemiologystudies[39-50].TheirsearchagreeswithallcurrentreviewsofglyphosateandIwillusetheirfindingsfromtheliteratureupuntil2015.TocoverfromJune2015tothepresent(April1,2017),Iusedtheirsearchingalgorithmandidentified117additionalpublishedstudies,noneofwhichwerenewepidemiologystudies.Thesesame12studieswillbeconsideredforuseinthisevaluation.Otherexpertswillbediscussingthestudiesaswellastheirstrengthsandtheirweaknesses;IwillfocusonusingtheresultsofthesestudiesinevaluatingcausalitysoIwillonlybrieflydescribeeachstudy.

Cantoretal.(1992)[39]didanin-personinterviewstudycomparing622whitemen,newlydiagnosedwithNHL,to1245population-basedcontrolsinIowaandMinnesota.Theyoriginallyidentified780cases,ofwhich694(89%)wereinterviewed.Afterpathologyreview,only622werefoundtohaveNHL,theremainingcaseshavingleukemiaorotherdiseases.Threedifferentsourcesofcontrolswereused,randomdigitdialing(76.7%responserate),HealthCareFinancingAdministrationrolls(79%response


7

rate)anddeceasedcontrolswitheligibleproxies(77%responserate).BothcasesandcontrolswerequestionedregardingtheiruseofagriculturalproductsincludingRoundup®andanyotherglyphosate-basedformulations.Fordeceasedorincompetentcontrols(184)andcases(numbernotgiven),proxyinterviewsweredonewithacloserelative.Whencasesinfarmerswerecomparedtocasesinnon-farmercontrols,26cases(outof266)and49controls(outof547)hadhandledherbicidescontainingglyphosateyieldinganoddsratio1(OR)of1.1(95%confidenceinterval0.7-1.9).Thisanalysiscontrolledforvitalstatus,age,state,cigarettesmokingstatus,familyhistoryoflymphopoieticcancer,high-riskoccupationsandhigh-riskexposuresinalogisticanalysis.Theauthorsnotedtherewas“minimalevidenceforconfoundingofresultsforanysinglepesticidebyexposuretopesticidesbelongingtootherchemicalfamilies.”Becausetheexposureisdeterminedbasedoninterviewsincasesandcontrols,thisstudyhasthepotentialforrecallbias2.However,theauthorsnotethatthebiascouldbothincreaseordecreasetheORbecauseofnon-differentialexposuremisclassification3becauseofdifficultiesinaccuraterecallofpastpesticideexposuresforbothcontrolsandtreatedindividuals.ThisstudywillnotbeincludedseparatelyintotheevaluationsinceitoverlapswithDeRoosetal.(2003)[43]

TwoadditionalstudiesconductedbyZahmetal.(1990)[51]inNebraskaandHoaretal.(1986)[52]inKansascollectedinformationonpesticideandherbicideuse,butdidnotreportspecificallyontheeffectsofglyphosate.DeRoosetal.(2003)[43]pooledthedatafromthesetwostudieswiththedatafromCantoretal.(1992)[39]toexaminepesticideexposuretoglyphosateinfarmingasriskfactorsforNHL.Thethreecase-controlstudies[39,51,52]hadslightlydifferentdesigns.ThedesignfortheMinnesotastudy[39]is

1Theoddsratio(OR)iscalculatedastheproportionofexposedcaseswithdiseasetoexposedcontrolsdividedbytheproportionofnon-exposedcasestonon-exposedcontrols.Forrarediseases,thisvalueapproximatesthepopulationriskratio(PRR)whichistheprobabilityofhavingthediseaseinexposedindividualsdividedbytheprobabilityofhavingthediseaseinnon-exposedindividuals.IfthePRRis1,thenthereisnodifferenceintheprobabilityofhavingthediseaseregardlessofyourexposure.ValuesofPRRgreaterthan1implytheriskishigherintheexposedpopulation.BecausetheORisanestimateofthePRRforrarediseases,itisusuallyaccompaniedbya95%confidenceintervalthatdescribestheprobablerangeoftheestimate.IftheORisgreaterthan1,thentheexposureisassociatedwiththedisease.Ifthelower95%confidenceboundfortheORisgreaterthan1,thisistypicallyusedtosaytheassociationisstatisticallysignificant.2Recallbiasoccurswhencasesaremorelikelytosaytheyareexposedtoglyphosatethancontrolsorwhencontrolsaremorelikelytosaytheyareexposedtoglyphosatethancases.Therecallmustbedifferentforthecasesthanthecontrolsforthistocauseabias;errorsinrecallingpastexposuresthathappenforbothcasesandcontrolswouldnotberecallbias.3Non-differentialexposuremisclassificationoccurswhentheprobabilityofanerrorindeterminingwhetheranindividualisexposedornotisthesameforbothcasesandcontrols.


8

provideddirectlyabove.InNebraska[51],thecaseswereidentifiedthroughtheNebraskaLymphomaStudyGroupandareahospitalsfor66countiesandincludedallwhitemenandwomendiagnosedwithNHLbetweenJuly1,1983andJune30,1986.Controlswereobtainedbyrandom-digitdialing,Medicarerecordsorstatemortalityfilesdependinguponageandvitalstatus.Allstudyparticipantswereoverage21andeventhoughthisstudyincludedafewwomen,theywereexcludedfromtheDeRoosetal.(2003)analysis.Theresponseratesforcasesandcontrolswere91%and87%respectively.InKansas[52],caseswererandomlysampledfromaregistryattheUniversityofKansasofwhitemen,overage21,diagnosedbetween1979and1981.Theresponseratesforcasesandcontrolswere96%and94%respectively.Controlswerepopulation-basedmatchedonageandvitalstatus.AsfortheNebraskastudy,controlsforlivecaseswereobtainedfromMedicarerecordsforcases65+andbyrandom-digitdialingforcases<65years;controlsfordeceasedpatientscamefromstatemortalityrecords.Theresultingpooledcase-controlstudyhad870casesand2569controls(foranalyzingtherelationshipbetweenglyphosateandNHL,therewereonly650casesand1933controlsfollowingexclusionofsubjectswithmissingdata).Foranyglyphosateexposure,therewere36exposedcasesand61exposedcontrolswithanOR(95%confidenceinterval)of2.1(1.1-4.0)inalogisticregressionanalysiscontrollingforallotherpesticidesreported,ageandstudysite.TheauthorsalsoanalyzedthedatausingaBayesianhierarchicalregressionanalysisyieldinganOR(95%confidenceinterval)of1.6(0.9-2.8)controllingforthesameparametersasthelogisticregression.Theyalsoconductedananalysisof“potentiallycarcinogenic”pesticideswhichincludedglyphosate.Whenjustoneofthesepesticideswasusedbysubjects,thelogisticregressionORwas1.6(0.8-3.1),twotofourpesticidesyieldedanORof2.7(0.7to10.8)andwhenmorethanfivewereused,theORwas25.9(1.5-450.2)inthelogisticregressionanalysisand1.1(0.8-1.7),1.3(0.7-2.3)and2.0(0.8-5.2)respectivelyfortheBayesiananalysis.Removingglyphosatefromthelistof“potentiallycarcinogenic”pesticidesyieldedequivalentORsof1.2foronepesticide,1.2fortwotofourpesticidesand1.1forfiveormorepesticides.Theauthorsnotethatthepositiveresultsseenintheirstudyarenotlikelyduetorecallbiassincetherewerefewassociationsseenoverthe47pesticidestheystudied.Also,althoughsomeofthepositiveresultscouldbeduetochance,theuseofthehierarchicalregressionanalysistheoreticallydecreasesthechanceoffalsepositivefindings.IntheKansasstudy[52],suppliersfor110subjectswithfarmingexperiencewereidentifiedandprovidedinformationonthesubjects’cropsandpesticidepurchases.Ingeneral,thesuppliersreportedlesspesticideusethanthesubjectsofthestudywithnoconsistentdifferencesinagreementratesbetweencasesandcontrols.Theagreementbetweensuppliersandsubjectsimprovedwhenpesticideuseduringthelast10yearswasconsidered.Thissupportsareducedroleofrecallbiasinthesestudiesandapossibleroleofnon-differentialexposuremisclassification.ThereducedORswhenusingtheBayesiananalysisascomparedtothelogisticregressionisnotsurprisingbecausetheauthorsusedanon-informativepriorratherthanalessconservativeprior.Inaddition,adjustmentfor47pesticidesisalsolikelytoreducethesignificanceoftheobservedORsforpesticidesthatareassociatedwithNHLasdemonstratedbytheanalysisof“potentiallycarcinogenic”pesticides(thismodelispossiblyover-parameterizedsinceit


9

includesover47dependentvariablesforonly36exposedcases;thiscansignificantlyreducetheORsandincreasetheconfidencebounds).Thispooledcase-controlstudyisthestrongeststudywithsufficientpower(3.8%ofsubjectsexposed)andwillbeincludedintheevaluationofcausation.

Leeetal.(2004)[44]pooleddatafromZahmetal.(1990)[51]andCantoretal.(1992)[39](previouslydescribed)toevaluatewhetherasthmaactsasaneffectmodifieroftheassociationbetweenglyphosateexposureandNHL.WomenwereincludedinthisanalysiswhereasDeRoosetal.(2003)[43]excludedwomen.ThefinalstudypublishedbyLeeincluded872casesand2336controlsofwhich45casesand132controlshadbeentoldbytheirdoctorstheyhadasthma.TheORofassociationbetweenglyphosateandNHLinnon-asthmaticswas1.4(0.98-2.1)and1.2(0.4-3.3)inasthmaticswhencontrollingforage,vitalstatusandstate(geographicallocation).ThisstudycompletelyoverlapswiththestudybyDeRoosetal.(2003)[43]withtheexceptionoftheinclusionofthefewwomeninthestudybyZahmetal.(1990)[51].Sincethisstudyonlylooksateffectmodificationduetoasthma,itdoesnotcontributetotheoverallevaluationofcausalityanditwillbeexcludedfromfurtherevaluations.

Nordstrometal.(1998)[40]conductedapopulation-basedcase-controlstudyofhairycellleukemia(HCL);asubtypeofB-cellNHL)inSwedenthatincludedanevaluationofexposurestoglyphosate.Thestudyincluded111menwithNHLreportedtotheSwedishCancerRegistrybetween1987and1992(withonepatientfrom1993accidentallyincluded).Controls(400intotal)weredrawnfromtheNationalPopulationRegistrymatchedforageandcountywiththecases.Theresponserateswere91%forcases(10refusedtoparticipateoutoftheoriginal121)and83%(84controlsrefusedtoparticipateoutof484selected).Almostallquestionnaireswereansweredbythesubjectofthestudy(4casesand5controlswereansweredbyproxies).ThestudyreportedanORforglyphosateexposureandHCLof3.1(0.8-12)controllingonlyforage.Thisstudyhadverylimitedpowerfordetectinganassociationbecausetherewereonlyfourcasesandfivecontrolswithglyphosateexposure(1.8%ofthetotalstudypopulation).Inaddition,becausetheyfailedtoadjustforotherexposures,thepotentialforconfoundinginthisstudyisgreaterthanthosepresentedpreviously.Theauthorsnotedthattheyattemptedtominimizerecallbiasbyonlyusinglivingcasesintheanalysis.Also,eventhoughmatchingwasperformedtoidentifythecontrols,thismatchingwasnotusedinthefinalanalysis.ThisstudywaslaterusedinapooledanalysisofHCLandNHL[42]andwillnotbeconsideredindependentlyintheevaluationforcausationbutwillbeusedinthecontextofthepooledanalysis.

HardellandEriksson(1999)[41]conductedapopulation-basedcase-controlstudyofallmalepatientsolderthan25yearsdiagnosedwithNHLbetween1987and1990inthefourmostnortherncountiesofSweden.Afterexcludingmisdiagnosedcases,theyincluded442casesofwhich404answeredtheirquestionnaire(mostbyproxy)foraresponserateof91%;192ofthesecasesweredeceased.Foreachlivingcase,twomalematchedcontrolswerechosenfromtheNationalPopulationRegistryandmatchedonageandcounty.Foreachdeceasedcase,twomalecontrolswerechosenfromtheNationalRegistryforCausesofDeath,matchedforageandyearofdeath.Theresponse


10

rateforthecontrolswas84%(741outof884identified).Studysubjectsweresentadetailedquestionnaireand,inmostcases,thiswassupplementedwithaphoneinterview.Acompleteworkinghistorywasobtainedwithquestionsregardingexposuretonumerouschemicalstoavoidafocusonpesticidesandorganicsolvents,thefocusofthestudy.Exposurewasdefinedasatleastonefulldayofexposuremorethanoneyearbeforediagnosis.Forglyphosateexposure,theauthorsidentifiedfourcasesandthreecontrolswithexposuresandaunivariateORof2.3(0.4-13).AmultivariateanalysisofbothglyphosateandphenoxyherbicidesproducedanORof5.8(0.6-54).Thestudyhaslimitedpowerfordetectinganeffectbecausetheexposurefrequencyisverylow(0.6%exposed).ThisstudywaslaterusedinapooledanalysisofHCLandNHL[42]andwillnotbeconsideredindependentlyintheevaluationforcausationbutwillbeusedinthecontextofthepooledanalysis.Hardelletal.(2002)[42]conductedapooledanalysisofNHLandHCLbycombiningthestudiesofNordstrometal.(1998)[40]andHardellandEriksson(1999)[41].Thisstudyfullyoverlapswiththeprevioustwostudies.Theanalysiscontrollingforage,study,countyandvitalstatusyieldedanORof3.04(1.08-8.52)basedoneightexposedcasesandeightexposedcontrols.AmoreextensiveanalysisadditionallycontrolledforotherpesticidesandyieldedasmallerORof1.85(0.55-6.20).AsforthestudybyDeRoosetal.(2003),theanalysismaybeover-parameterized(morethaneightdependentvariableswithonlyeightexposedcases)whichcouldleadtoareductionintheORsandlargerconfidencebounds.Evenwiththepooleddata,Hardelletal.(2002)hadlimitedpowertodetectaneffectbecausetheexposurefrequencyforcasesandcontrolswasverylow(1%exposed).Thisstudyisavalidcase-controlstudyandwillbeusedintheevaluationofcausality.

Inalaterstudy,Erikssonetal.(2008)[46]conductedapopulation-basedcase-controlstudywherecaseswereidentifiedasNHLpatientsaged18-74yearsdiagnosedinfourmajorhospitalsinSwedenfromDecember1,1999untilApril30,2002.Intotal,995caseswereidentifiedasmatchingthestudyparameterswith910(91%)answeringthequestionnaireshortlyafterdiagnosis.Allcaseswereclassifiedintosubgroupswith810B-cell,53T-cell,and38unspecifiedlymphomas.Controls(1,108)wererandomlyselectedfromthepopulationregistryandmatchedonhealthservice,region,sexandageandinterviewedinseveralperiodsduringtheconductofthestudy;1,016controlsrespondedtothequestionnaire(92%responserate).Studysubjectsweresentadetailedquestionnaireand,inmanycases,aphoneinterviewfollowed.Exposurewasdefinedasatleastonefulldayofexposuremorethanoneyearbeforediagnosis.Theunivariateanalysis,adjustingforage,sexandyearofdiagnosis(cases)orenrollment(control)yieldedanORof2.02(1.10-3.71)basedon29exposedcasesand18exposedcontrols.Whencasesandcontrolsweredividedintothosewith≤10daysperyearexposureandthosewith>10daysperyearexposure,theORswere1.69(0.70-4.07)and2.36(1.04-5.37)respectively.WhendiagnosesweregroupedintovarioussubtypesofNHL,theresultsdidnotchangedramaticallyexceptforsmalllymphocyticlymphomaandchroniclymphocyticlymphomawhichshowedanincreasedORof3.35(1.42-7.89).Amultivariateanalysisofglyphosatecontrollingforotheragentswithstatistically


11

increasedoddsratiosand/oroddsratiosgreaterthan1.5yieldedanORof1.51(0.77-2.94).Inasimilaranalysistothemultivariateanalysis,latencyperiodsofonetotenyearsshowedanORof1.11(0.24-5.08)and>10yearshadanORof2.26(1.16-4.40).ThisstudywasmuchlargerthanthepreviousSwedishstudies(2.3%exposed)and,althoughtheremayhavebeenconfoundingfromotherpesticides,thiswasaddressedinthemultivariateanalysisandthelatencyanalysis.Thisstudyisavalidcase-controlstudyandwillbeusedintheevaluationofcausality.

McDuffieetal.(2001)[50]recruitedincidencecasesofNHLinmen19yearsorolderfromsixCanadianprovinceswithafirstdiagnosisbetweenSeptember1,1991andDecember31,1994.EachprovincialCancerRegistryor,inthecaseofQuebec,hospital,hadatargetnumberofcasesandendedrecruitmentwhenthecasenumberwasreached.Controlsweremen19yearsorolderselectedatrandomfromprovincialhealthinsurancerecords,computerizedtelephonelistingsorvoterregistrationlists,dependingupontheprovince.Casesandcontrolsweresentquestionnaireswithsurrogatesineligibletoanswerthequestionnairesfordeceasedcasesorcontrols.Eachsubjectwhoreported10hoursperyearormoreofpesticideexposureandarandomsampleof15%whoreportedlessexposurewereinterviewedbytelephonetoobtaindetailsonpesticideuse.ApilotstudywasconductedtoobtainanimprovedversionofthetelephoneinterviewquestionnaireusedbyHoaretal.(1986)[52]andZahmetal.(1990)[51]thatwouldprovideaccuratepesticideexposureassessmentintheformofascreeningquestionnaireandatelephoneinterviewquestionnaire.Thiswasfollowedbyavalidationstudy(27farmers)wherethefinalquestionnairesusedtoscreenandincludepotentialcasesandcontrolswereadministeredandtheanswersregardingpesticideusageshowedexcellentconcordancewithpurchasesthroughtheirlocalagrochemicalsupplier.Thescreeningquestionnairewasreturnedby517casesofNHL(67.1%responserate)and1506controls(48%responserate).Followinganalysisofthescreeningquestionnaire,thetelephoneinterviewwasadministeredto179casesand456controlstoobtainmoredetailedexposureinformation.TheORforglyphosateexposureandNHLwas1.26(0.87-1.80)stratifiedbyagegroupandprovinceofresidenceandtheORwas1.20(0.83-1.74)whentheanalysisalsocontrolledforsignificantmedicalvariables(51exposedcasesand133exposedcontrols).Anexposure-responseevaluationwasperformedwheretheORforexposurebetweenzerototwodaysperyearwas1.0(0.63-1.57)andforgreaterthantwodaysperyearwas2.12(1.20-3.73)withthelattergrouphaving23exposedcasesand36exposedcontrols.Thisstudyhadexcellentsamplesizeandpower(8.1%ofsubjectsexposed),butalowresponseratetothescreeningquestionnaire.Also,byadjustingforsignificantmedicalvariables,thisstudyruledoutmanyconfoundersbutdidnotadjustforotherpesticideexposures.Theefforttovalidatetherecallofpesticideusageforfarmerssupportsalackofrecallbiasinthestudy.Thisstudyisavalidcase-controlstudyandwillbeusedintheevaluationofcausality.

Hohenadeletal.(2011)[48]re-analyzedthedataofMcDuffieetal.(2001)[50]tospecificallyinvestigatetheimpactofexposuretomultiplepesticidesonNHL.FourcasesofNHLwereexcludedfromthisevaluationfollowingapathologyreview.Theyreportedassociationswiththeuseofglyphosatewithandwithoutmalathionbutnotwith


12

glyphosateoverall.TheORforglyphosate(everused)withoutmalathion(everused)was0.92(0.54-1.55)andtheORforglyphosate(everused)withmalathion(everused)was2.1(1.31-3.37).ChangandDelzell(2016)[38]combinedtheORsfromtheglyphosateonlyanalysiswiththeglyphosateandmalathionanalysesusingrandom-effectsmeta-analysistogetacombinedORforglyphosateof1.4(0.62-3.15).Thisstudywasspecificallytargetedtointeractionsofvariouspesticidesanddoesnotsubstantivelycontributetoanevaluationofglyphosate.SinceitisarefinedanalysisofMcDuffieetal.(2001)[50],itwillbeincludedintheevaluationofcausationonlyinthecontextofthecombinedanalysisprovidedbyChangandDelzell(2016).

Orsietal.(2009)[47]conductedahospital-basedcase-controlstudyofmenandwomendiagnosedwithlymphoidneoplasmsinfivehospitalsinFrancebetween2000and2004whowereaged20-75years(theabstractgivestheagerangeas18-75years).Alldiagnoseswerecytologicallyorhistologicallyconfirmed.Theevaluationonlyincludedmenandquestionnaires/interviewswerecompletedby491cases(95.7%responserate)whichincluded244caseswithNHL.Controlswerepatientsinthesamehospital(mostlyorthopedicorrheumatologicalpatients)withnopriorhistoryoflymphoidneoplasmsandexcludingpatientsadmittedtothehospitalforcanceroradiseasedirectlyrelatedtooccupation,smokingoralcoholabuse.Thecontrolswerematchedtocasesbyhospitalandage.Ofthe501candidatecontrols,456participated(91%response).Exposurewasevaluateddifferentlyforsubjectswhohadnon-occupationalexposuresfromthosewhohadoccupationalexposures.Forboth,thesubjectshadtofilloutaquestionnaire/interviewonoccupationsandhomegardeningpesticideexposures.Forthosewhohadworkedprofessionallyasfarmersorgardenersforatleast6months,aspecificagriculturaloccupationalquestionnaire/interviewwasadministeredandexposurewasdeterminedonthebasisofthisextradata.TheORforoccupationaluseofglyphosateandNHLwas1.0(0.5-2.2)with12exposedcasesand24exposedcontrolsstratifiedbyageandcentercategory.AfurtheranalysiswasdonebyindividualsubtypesofNHLwithanORof1.0(0.3-2.7)fordiffuselargecelllymphoma,1.4(0.4-5.2)forfollicularlymphoma,0.4(0.1-1.8)forchroniclymphocyticleukemia(CLL)and1.8(0.3-9.3)forHCL.Noseparateanalysisofnon-occupationaluseofglyphosatewasprovided,nordoesitseemspecificdataonglyphosateusagewasascertainedforsubjectswhowerenotprofessionalfarmersorgardeners.Thiscouldleadtonon-differentialmisclassificationofexposurewhichcouldreducetheORsofthestudy.Barringthis,thesamplesizewassufficienttodetectaneffect(5.3%withoccupationalexposure)andthisstudywillbeincludedintheevaluationofcausality.

Coccoetal.(2013)[49]evaluateddatafromamulti-centercase-controlstudyoflymphoidneoplasmsinsixEuropeancountriesfrom1998to2004.Casesincludedonlyadultpatientsdiagnosedwithlymphomaduringthestudyperioddrawnfromparticipatingcenters.Controlswereeitherselectedbysamplingfromthegeneralpopulationonsex,agegroup,andresidencearea(Germany,Italy),orfromhospitalcontrolsmatchedtothepatientexcludingpatientswithcancer,infectiousdiseases,andimmunodeficiencydiseases(CzechRepublic,France,Ireland,Spain).Thestudyincluded2348lymphomacases(88%participation)and2462controls(81%responserateinhospital-basedcontrolsand52%inpopulation-basedcontrols).Exposureswerederivedusingan


13

occupationalexposurematrixdevelopedbyindustrialhygienistsandoccupationalexpertsfromtheresearchcenters.Only35individuals(casesandcontrolsnotbrokenout)inthestudywereexposedtocarbamates(glyphosatewasgroupedwiththecarbamates).NoresultswereprovidedforNHLandtheonlyORprovidedforglyphosatewasforB-celllymphomawheretheORwas3.1(0.6-17.1)basedonfourexposedcasesandtwoexposedcontrols.Noinformationwasprovidedonthetotalnumberofcasesforeachtypeoflymphomaevaluated.ThisstudyhasverylimitedpowertoevaluateanassociationbetweenNHLandglyphosateandprovidesonlyinformationonB-celllymphomaswithveryfewexposedcasesandcontrols.Ashasbeendonebymostresearchersevaluatingthesedata,thisstudywillreceiveverylittleweightintheevaluationofcausality.

DeRoosetal.(2005)[45]reportedresultsontheassociationofglyphosateandcancerincidencefromtheAgriculturalHealthStudy(AHS),aprospectivecohortstudyinIowaandNorthCarolina,whichincluded57,311privateandcommercialapplicatorswhowerelicensedtoapplyrestricted-usepesticidesatthetimeofenrollment.Recruitmentoccurredbetween1993and1997andcohortmemberswerematchedtocancerregistryfilestoidentifycasesandtheNationalDeathIndex(1999)toascertainvitalstatus.Incidentcancerswereidentifiedfromthedateonenrollmentuntil31December,2001,withtheaveragefollow-uptimebeing6.7years.Comprehensiveusedatawasobtainedbyself-administeredquestionnairefor22pesticides,ever/neverusefor28additionalpesticides,andgeneralinformationonworkpractices.Applicatorsweregivenasecondself-administeredquestionnaireonoccupationalexposuresandlifestylefactors.Theyusedthreeexposuremetricsintheiranalyses:a)everpersonallymixedorappliedpesticidescontainingglyphosate;b)cumulativeexposuredaysofuseofglyphosate(yearsofusetimesdaysperyear);andc)intensityweightedcumulativeexposuredays(yearsofusetimesdaysperyeartimesintensityofuse).Personswhosefirstprimarytumoroccurredbeforethetimeofenrollment(1074)wereexcludedfromtheanalysisaswerethosewhowerelosttofollow-up(298),didnotprovideageinformation(7)orinformationonglyphosateuse(1678)leaving54,315subjectsforinclusion.Therewere92cohortmemberswithadiagnosisofNHLduringthestudyperiodofwhich77.2%hadeverusedglyphosateresultinginarateratio4(RR)of1.2(0.7-1.9)whencontrollingforageandanRRof1.1(0.7-1.9)whencontrollingforage,lifestylefactors,demographicsandfiveotherpesticidesforwhichcumulative-exposure-dayvariablesweremosthighlyassociatedwithglyphosatecumulative-exposure-days(2,4-D,alachlor,atrazine,metalochlor,andtrifluralin)or,forchemicalswithonlyever/neverexposureinformationthatweremosthighlyassociatedwithglyphosateever/neveruse(benomyl,maneb,paraquat,carbarylanddiazinon).WhencumulativeexposuredaysinexposedindividualsaredividedintotertilesandRRsexaminedusingthelowestexposedtertileas

4Therateratio(RR)isestimatedastheincidenceintheexposedpopulationdividedbytheincidenceintheunexposedpopulation.Incidenceiscalculatedasthenumberofeventsinafixedperiodoftimedividedbythepersonyearsatrisk.UnliketheOR,theRRdoesnotrequiretheassumptionofararediseasetoserveasagoodestimateofthepopulationriskratio(PRR).


14

thereferencegroup,theRRsdropwithvaluesof0.7(0.4-1.4)and0.9(0.5-1.6)fortertiles2and3respectivelycontrollingfordemographicandlifestylefactorsandotherpesticides(30,699subjects).Whenintensity-weightedexposuredaysareexaminedagainusingexposedtertile1asthereferencegroup,theRRsdropwithvaluesof0.6(0.3-1.1)and0.8(0.5-1.4)fortertiles2and3intensity-weightedexposuredaysrespectivelycontrollingfordemographicandlifestylefactorsandotherpesticides(30,699subjects).Analysesarenotshownfortheevaluationoftheexposedtertilesagainstneverexposedbecausetheauthorsfeltthatneverexposedandexposedsubjectsdifferedintermsofsocio-economicfactorsandotherexposureslikesmoking[45].

Thisisatypicalcohortstudy,buthassomelimitationsintermsofitsinterpretation.Themajority(75.5%)ofsubjectsinthecohortreportedhavingeverpersonallymixedorappliedproductscontainingglyphosateandwascomposedprimarilyofmale,middle-aged,privateapplicators.Forglyphosate,reliabilityoftheanswersbysubjectsontheuseofglyphosatebetweenthefirstandsecondquestionnairewereevaluatedintheAHS[53]:82%agreementforwhethertheyhadevermixedorappliedglyphosate,53%agreementonyearsmixedorapplied,and62%agreementondaysperyearmixedorappliedand62%agreementondecadefirstapplied.Theysawnodifferencesinoverversusunderreportingbetweenthetwoquestionnairessuggestingthiscouldleadtonon-differentialexposurebiasandreducetheRRsinthisstudy.Anotherweakness,notedbytheauthors,isthatthesmallnumberofincidentcasesduringfollow-upperiodhinderedpreciseeffectestimates.Also,thehighfrequencyofexposuretomanypesticides(e.g.73.8%wereexposedto2,4-D)meanssubjectsunexposedtoglyphosatewerelikelytobeexposedtootheragentsthatmayalsoinduceNHL,reducingtheRRs.Also,asnotedbytheEPA’sFIFRAScienceAdvisoryPanel(SAP)[54]intheirreviewoftheEPA’sissuepaperonthecarcinogenicityofglyphosateandasnotedinacritique[55]oftheEuropeanFoodSafetyAgency’sriskassessmentforglyphosate,thefollow-uptimeinthiscohortstudymaynotbelongenoughtoproduceasufficientsamplesizeforevaluationoftheassociationbetweenNHLandglyphosate.Likeotherstudies,thisstudyhasfewexposedcasesandcontrols,buttheauthorsadjusttheiranalysisformanyotherpesticideswhichcouldreduceORsandincreaseconfidenceboundslimitingtheabilityofthestudytoshowpositiveresults.Thisstudycouldalsosufferfromasurvivalbiasbecausepesticideapplicatorswererecruitedascaseparticipantsaftertheirexposurehadbegunandthosewithacancerpriortoenrollmentwereexcluded.

Thisstudywillbeincludedintheevaluationofcausality.

ConsistencyofAssociationsHill(1965)[36]definesconsistencyastheanswer“yes”tothequestion“Hasitrepeatedlybeenobservedbydifferentpersons,indifferentplaces,circumstancesandtimes?”Forthesestudies,theanswerisindeedyes.

Ifthepopulationrelativerisk(PRR)foranassociationofglyphosatewithNHLwereequalto1(noeffect),thenonewouldexpectveryfewstatisticallysignificantresultsinmultiplestudiesandthatabouthalfofthestudieswouldhaveORsorRRsbelowone


15

andhalfaboveone.AsnotedbyboththeIARCMonograph112(2015)[56]andbyChangandDelzell(2016)[38],whencomparingstudies,themostreasonablecomparisonistousethemost-fully-adjustedriskestimates.Iwillmostlylimitmycommentstothesemost-fully-adjustedriskestimates.

Consistencyoftheassociationsacrossseveralepidemiologystudiesisnotsimplyamatterofseeinghowmanywerestatisticallysignificantandhowmanywerenotbutmustalsoaddresstheconsistencyofthedirectionoftheresponses.Figure1showsaforestplotofallORsandRRsfromtheepidemiologystudiesdiscussedpreviously.EachhorizontallineintheforestplotshowsthemeanestimateoftheOR/RRasablacksquareandthe95%confidenceintervalaroundthisestimateaswhiskersextendingleftandrightfromtheblacksquare.

ThefirstobviousconclusiontobedrawnfromFigure1isthatallofthemeanOR/RRestimates(blacksquares)areconsistently≥1.Thisimpliesthatallofthestudiesarepointinginthesamedirectiontowardapositiveeffect.Intheirmeta-analyses,SchinasiandLeon(2014)[57],IARC(2015)[56]andChangandDelzell(2016)[38]allidentified6papers(highlightedinredinFigure1)asbeingthemostreliableforevaluationoftheabilityforglyphosatetoinduceNHLinpeople:McDuffieetal.(2001)[50],Hardelletal.(2002)[42],DeRoosetal.(2003)[43]and(2005)[45],Erikssonetal.(2008)[46]andOrsietal.(2009)[47].Iwillrefertothesepapersasthesixcoreepidemiologystudies.Asnotedabove,ifthetrueunderlyingriskratiowas1(noeffect),youwouldexpectabouthalfofthefindingstobebelow1andhalftobeequalto1orgreater.Usingonlytheresultsfromthe6corestudies,youcanseethatallare≥1;theprobabilityofthishappeningis(0.5)6or0.016,stronglysuggestingthestudiesdonotagreewithanunderlyingPRR=1andthattheyconsistentlysupportapositiveeffect.

Asecondwayinwhichconsistencycanbeevaluatedistocombinetheindividualstudiesusingmeta-analysistoobtainacombinedanalysisusingboththeORsandtheRR(CRR)andtestforheterogeneityinthestudies.Themeta-analysisdonebyChangandDelzell(2016)includesthesameanalysisasthatdonebytheIARC(2015)andisanimprovementoverSchinasiandLeon(2014),soIwillfocusmycommentsonusingtheChangandDelzell(2016)meta-analysis.ChangandDelzell(2016)didfourseparatemeta-analysesontheglyphosateepidemiologystudiesusingtwodifferentmethods(random-effectsandfixed-effectsmodels).Intheirfirstanalysis(model1)5,theycombinedthemost-fully-adjustedriskestimatesfromthesixcorestudiestoyieldaCRRof1.27(1.01-1.59)forbothrandom-effectsandfixed-effectsmodelssupportinganassociationbetweenNHLandglyphosateexposureinthesestudies.Inasecondanalysis(model2),theyreplacetheresultsoftheBayesiananalysisinDeRoosetal.(2003)withtheresultsofthelogisticregressionanalysisandgetthesameCRRof1.30(1.03-1.64)forbothrandom-effectsandfixed-effectsmodels.Inathirdanalysis(model3),theyreplacefrommodel1theMcDuffieetal.(2001)resultsinwithacombinedmeta-

5ChangandDelzell(2016)providedonlyonesignificantdigittotherightofthedecimalpointintheirconfidencebounds;theEPASAP(2017)re-calculatedmodels1-4ofChangandDelzell(2016)toprovidetwosignificantdigits–thesearepresentedhere.


16

analyticresulttheyderivedfromanalysesbyHohenadeletal.(2011)(thisstudyreanalyzedthesamedataasMcDuffieetal.(2001),splittingresultsbetweenasthmaticsandnon-asthmatics)resultinginaCRRof1.32(1.00-1.73)forbothrandom-effectsandfixed-effectsmodels.Finally,inafourthanalysis(model4),theyusemodel3butreplacedtheBayesiananalysisinDeRoosetal.(2003)withthelogisticregressionanalysisyieldingaCRRof1.37(1.04-1.82)forbothrandom-effectsandfixed-effectsmodels.Inessence,noneofthedifferentmeta-analysesrejectedthenotionofacombined,statisticallysignificantpositiveeffect.

Figure1:OddsRatiosandRateRatiosfromthemost-fully-adjustedriskestimatesfromselectedepidemiologystudiesandfromthemeta-analysesofChangandDelzell(2016)[38].“RR”referstotheORorRRfromthestudy,“Lower”referstothe95%lowerbound,“Upper”tothe95%upperboundand“Weight”referstotheweightappliedtothatspecificstudyinModel1ofthemeta-analysis(Table3inChangandDelzell).ForDeRoosetal.(2003),thefirstrowisfortheBayesianmodelanalysisandthesecondrow,labelled“logisticregression”isfromthelogisticmodelanalysis.

Asstatedabove,anotherwaytoevaluateconsistencyintheepidemiologicaldatawouldbetoevaluatetheheterogeneityinthestudies.Heterogeneitymaybeduetodifferencesinparticipants,outcomes,exposuremetrics,methodsforquestioningstudysubjects,sexofthesubjects,etc.ChangandDelzell(2016)formallytestedforheterogeneityoftheresponsesfromthesixcorestudiesusingCochran’sQstatisticandtheI2statistic[58].Formodels1to4,thep-valuesfromCochran’sQtestare0.84,0.59,0.85,and0.63respectively(typicallyyourejecttheconceptofhomogenousstudiesinfavorofheterogeneousstudiesifp<0.10).TheI2statisticforallfourmodelsare0.0%(valuesforI2canrangefrom0-100%withconcernforheterogeneityabove50%).Thefactthatthefixed-effectsmodelsandrandom-effectsmodelsgavethesameresultsalsosupportsalackofheterogeneityinthedata.Thereisnoindicationofheterogeneityinthesesixcorestudies.Lackofheterogeneitysupportstheinterpretationofthemeta-analysesasshowingapositiveassociationandstrongconsistencyofthefindingsacrossthesixcorestudies.


17

ChangandDelzell(2016)alsoevaluatedtheassociationbetweensubtypesofNHLandglyphosateexposurewherepossible.ForB-celllymphomas,theycombinedtheresultsofErikssonetal.(2008)[46]withthoseofCoccoetal.(2013)[49]andsawaCRR(random-effectsandfixed-effects)of2.0(1.1-3.6)withanI2of0andaCochran’sQtestp-valueof0.58.FordiffuselargeB-celllymphomas,theycombinedtheresultsofErikssonetal.(2008)[46]withthoseofOrsietal.(2009)[47]andsawaCRR(random-effectsandfixed-effects)of1.1(0.5-2.3)withanI2of0andaCochran’sQtestp-valueof0.79.Forcombinedchroniclymphocyticleukemiaandsmalllymphocyticlymphoma,theycombinedtheresultsofErikssonetal.(2008)[46]withthoseofOrsietal.(2009)[47]andsawaCRRusingtherandom-effectsmodelof1.3(0.2-10)andforthefixedeffectsmodel1.9(0.9-4.0)withanI2of83.7%andaCochran’sQtestp-valueof0.01.Forfollicularlymphomas,theycombinedtheresultsofErikssonetal.(2008)[46]withthoseofOrsietal.(2009)[47]andsawaCRR(random-effectsandfixed-effects)of1.7(0.7-3.9)withanI2of0andaCochran’sQtestp-valueof0.73.Andfinally,forHCL,theycombinedtheresultsofNordstrometal.(1998)[40]withthoseofOrsietal.(2009)[47]andsawaCRR(random-effectsandfixed-effects)of2.5(0.9-7.3)withanI2of0andaCochran’sQtestp-valueof0.63.Thesesubtypeanalysesarebaseduponsmallnumbersofcasesandonlytwostudiesmakingthemunreliable,whenconsideredindividually,toaddressthequestionofconsistencyinthedata.However,whentheyarecombinedwiththeresultsforthemeta-analysesofthecorestudiesofNHL,thesestudiesaddsupporttotheconclusionthatthesedataareconsistent.

ChangandDelzell(2016)alsoperformedasensitivityanalysisbyonlydoingmeta-analysesonstudieswithsimilarcharacteristics.Usingonlythefivecase-controlstudies,theCRRwas1.3(1.0-1.7).Breakingthemintothetypeofcontrolused,therewerefourstudiesusingpopulationcontrolswithaCRRof1.4(1.0-1.8).TherewerefourstudieswithmalesonlywithaCRRof1.3(1.0-1.7)andtwostudieswithmalesandfemaleswithaCRRof1.2(0.8-1.8).ThreestudiesweredoneinNorthAmericawithaCRRof1.2(1.0-1.6),threeinEuropewithaCRRof1.3(0.8-2.1);twoofthethreestudieswereinSwedenwithaCRRof1.6(0.9-2.8).AlloftheresultingmetaCRRswerethesameforthefixed-effectsmodelandtherandom-effectsmodel.ThissensitivityanalysisshowsthattheresultsdonotdiffersignificantlyfromthemainCRRforthesixcorestudiescombinedaddingsupporttothefindingsbeingconsistentacrossthedifferentstudies.

Incase-controlstudies,selectionbiasariseswhenthereasonscasesandcontrolschoosetoparticipateinthestudycouldleadtosystematicbiasesthatmightresultinapositiveornegativefindingindependentoftheexposurebeingstudied.Forexample,ifcaseswithexposurearemorelikelytoparticipatethancontrolswithexposure,theresultwouldbehigherORvalues;however,thisdifferencehastobedifferentialandnotsimplyadifferenceinparticipationrates.Itispossiblethatinafewofthesestudies,themethodbywhichcontrolswereselectedcouldcontributetoselectionbiasthatmightleadtoincreasedORs.However,giventhediversetypesofcasesandcontrolsusedinthefivecorecase-controlstudies,thisisunlikelytoexplaintheconsistentfindingsseenfromthesestudies.Itisalsopossiblethatthelackofcompletedataoncasesversuscontrolscouldresultinselectionbiasifthereasonsfornotcompletingthequestionnaire/interviewaredifferentbetweencasesandcontrolsandrelatesto


18

exposure.Thereisnoindicationofthistypeofselectionbiasinthesereports,andthisisunlikelytoexplaintheconsistencyseeninthesedata.

ExposuremisclassificationcanleadtoincreasesordecreasesintheORorRRvaluesseeninbothcase-controlandcohortstudies.Forexample,incase-controlstudies,ifcasesaremorelikelytosaytheywereexposedtoglyphosatethancontrols,thiswouldinflatetheORvalues;thisisonetypeofrecallbias.Thistypeofbiasislesslikelyincohortstudies.Inallsixofthecorestudies,thisissuewasdiscussedbytheauthors.Ineverycase,theyconcludedtherewasboundtobesomeexposuremisclassification,butthatitwasmostlikelynon-differential,meaningthatthemisclassificationwasrandom;thiswouldlikelyreducetheOR/RRsseeninthestudiesratherthanincreasethem.

ConfoundingoccurswhenthereisanexposureorsomeotherfactorthatistightlyassociatedwithbothglyphosateexposureandNHLdiagnosisthat,ifcontrolledfor,couldexplaintheresults.Themostlikelysourceofconfoundinginthesestudieswouldbeexposurestootherpesticides.Four[42,43,45,46]ofthesixcorestudiescontrolledforexposuretootherpesticidesandsawbasicallythesamefindingsastheothertwostudies.AnotherconcernforconfoundingwouldbeifthecaseshadimmunedeficienciesthatcouldbelinkedtoNHL;inallofthecase-controlstudies,suchcaseswereexcluded.Finally,otheragriculturalexposures(e.g.animals,otherchemicals,infectiousagents)couldbecorrelatedwithglyphosateexposureandmaybelinkedtoNHL;noneofthestudiescontrolledforthesefactors.However,notallexposedcaseswerefarmers;ifconfoundingviaotheragriculturalexposuresisoccurring,itisnotpossibletodeterminethemagnitudeordirectionofsuchaneffectfromthesedata.

Inconclusion,wehavesixcoreepidemiologystudiesdoneontwodifferentcontinentsbyfourdifferentresearchgroupsusingdifferentdesigns,questionnairesandstudypopulationsthatarehighlyconsistentwithnoobviousbiasorconfoundingthatwouldexplaintheresults.Thereisaconsistencyofassociationsacrossthesixcorestudies.

StrengthoftheAssociationseeninHumanEpidemiologicalStudies

Toexplainstrengthofassociation,Hill(1965)givestheclassicexampleofJohnSnowandthecholeraepidemicof1855wheretheriskratioofdyingifyoudrankwaterfromtheSouthwarkandVauxhallCompany(pollutedbysewage)comparedtodrinkingfromtheLambethCompanywater(sewagefree)was14.Yet,forthesixcorestudies,theOR/RRrangesfrom1.0to1.85forthemost-fully-adjustedriskestimatesandto2.1ifyouincludethefullyadjustedriskestimatefromDeRoosetal.(2003)[45]usinglogisticregression.ThesearemoderateOR/RRestimatesmakingitconceivabletheyareindividuallyduetoeitherchanceorbias.Thus,withtheexceptionofthelogisticregressionanalysisinDeRoosetal.(2003)[45],noneofthecorestudiesdemonstratelarge,preciserisksasenvisionedbyHill(2016)[36].However,Hill(1965)wasnotexpressinghimselfinstatisticaltermswherethesignificanceofanassociationisdependentupontheprecisionoftheobservations.IfthestatisticalvariationaroundanOR/RRestimateislargerelativetotheestimateitself,theestimateisnotveryprecise


19

andgenerallywouldnotbestatisticallysignificant.TheresultfromthestudybyHardellandEriksson(1999)showninFigure1isanexampleofanestimatewithverylargestatisticalvariation.Ontheotherhand,averysmall(invalue),preciseORorRRestimatecouldbestatisticallysignificantandproveimportantindecidingcausation.Themeta-analysesshowninFigure1alldemonstrateestimatesofOR/RRthataresignificantlydifferentfrom1rejectingtheconceptthattheoverallassociationisduetochance.ThestatisticallysignificantestimateoftheOR/RRforB-celllymphomasinthemeta-analysissupportthisfindingaswell.

Insummary,wehavesixcoreepidemiologystudiesthatallshowapproximatelythesame,modestincreaseinOR/RRthat,whencombined,demonstrateasignificantstrengthofassociation.Thereisastrongassociationacrossthesixcorestudies

BiologicalPlausibility

Therangeofdataonecanusetodeterminebiologicalplausibilityisquitediverseandcanbeexceptionallycomplicated.Forsimplicity,itcanbedividedintothetypesofassaysthatcanbeusedinthisevaluation:animalcancerbioassays,toxicokineticstudies,studiesfromaccidentalexposuresinhumans,andstudiesofspecificbiologicalmechanismsinanimalsorcellsderivedfromhumansoranimals.Animalcancerbioassaysareintendedtotestwhetherglyphosatecancausecancersinmammals,thussupportingtheconceptthatthechemicalcouldcausecancerinhumans.Toxicokineticstudiesprovideinsightintothedegreetowhichglyphosateisabsorbedbyhumans,distributedtovariousorgansinthebody,whathappenstothechemicalonceitisinthebody(metabolism),and,finally,howitiseliminatedfromthebody.Studiesfromaccidentalexposuresinhumanscanprovidesomeinformationontheeffectsofglyphosatethroughchangesinthechemistryandcellularstructureofhumanblood.Studiesofbiologicalmechanismsaregenerallyaddressingwhateffectsthechemicalmayhaveonhumanandanimalcellsundercontrolled,laboratoryconditions.Someofthestudiesinthissectionweredonewithtechnicalgrade(virtuallypure)glyphosateandsomewiththeglyphosateformulationsthathumansencounterinoccupationalandenvironmentalsettings.IwillsummarizetheliteratureineachoftheseareasandofferanopiniontotheirsupportofbiologicalplausibilityofNHLinhumans.

AnimalCancerBioassays

Typicalanimalcancerbioassayswillexposeanimals(generallyratsormice)toachemicalforasubstantialproportionoftheanimal’slife(generally2years)thenkilltheanimalandexamineitsorgansandtissuesfortumors.Thereareguidelinesonhowtoconductandanalyzethesestudies.Typically,chemicalregistrantsconductcancerbioassaysforpesticideapprovalpursuanttoguidelinesdevelopedundertheguidanceoftheOrganizationforEconomicCooperationandDevelopment(OECD[59]).Othergroups[30,33,34]provideguidanceonhowtoanalyzethesestudiesbaseduponmethodologypapersfromthepublishedliterature.Thesestudiesareconductedinawaythatcontrolsforeverythingintheanimal’senvironment(e.g.,foodtype,waterquality,howoftentheanimalsarehandled)leavingonlytheexposuretoexplain


20

differencesintumorformationbetweencontrolandexposedanimals.Eventhen,non-cancerendpointscanalsobemodifiedbythechemicalandthesemayhaveanimpactontumorratesintheanimals(e.g.,survival,deathfromsomeothertoxiceffectofthechemical);thesemustbeaccountedforwhenreachingconclusionsfromthestudy.

Studiesgenerallyusefourgroupsofanimals,onegroupreceivingnoexposure(control)andtheremainingthreegroupsaretestanimals,witheachgroupreceivingdifferentdoseexposurestothechemical[60].Dosesgenerallyabovehumanexperienceareusedinanimalcarcinogenicitystudiesbecauseonlyrelativelysmallnumbersofanimalsarebeingusedtoevaluateriskforalargehumanpopulationandbecauseeventhebestknownhumancarcinogensdonotcausecancerinlargefractions(say20%)ofthehumanpopulation.Thebasicunderlyingpremiseofthisdesignconsiderationisthat,asthedoseincreases,sodoestheriskofgettingatumor.Byexposinganimalstothehighestdosepossible,youincreasetheabilityofthestudytoidentifyariskifoneispresent.However,onemustbecarefulnottouseadosethatissohighitwillcausecancersbyprocessesthatwouldneverworkatlowerdoses.Toavoidthis,studiesaredesignedaroundamaximumtolerateddose(MTD)orlimitdose.Thisdoseisgenerallydeterminedbaseduponasubchronicstudy(90days)inthesameanimalsandisusuallythemaximumdosethatcanbetoleratedbytheanimalswithoutanysignsofsignificanttoxicityintheexposedanimals(e.g.,weightloss,tissuedamage).TheOECDandEPAprovideguidelines[33,59]onhowtochoosethistopdose.Theseguidelinesareingeneralagreementwiththescientificliterature[60].

Theguidelinesalsoaddressthemethodsbywhichthedatashouldbeanalyzed.Forexample,theEPAguidelines[61]statethat:

“AtrendtestsuchastheCochran-Armitagetest(SnedecorandCochran,1967)askswhethertheresultsinalldosegroupstogetherincreaseasdoseincreases.ApairwisecomparisontestsuchastheFisherexacttest(Fisher,1950)askswhetheranincidenceinonedosegroupisincreasedoverthatofthecontrolgroup.Byconvention,forbothtestsastatisticallysignificantcomparisonisoneforwhichpislessthan0.05thattheincreasedincidenceisduetochance.Significanceineitherkindoftestissufficienttorejectthehypothesisthatchanceaccountsfortheresult.”

Infact,mostguidelinesandpeer-reviewedpublicationscometothesameconclusion[30,59,60,62]onwhatteststouse,asdidEPA’sFIFRAScientificAdvisoryPanel(SAP)intheirreviewoftheEPA’sissuepaperofthecarcinogenicityofglyphosate[54].TheUSNationalToxicologyProgram(NTP)usesbothatrendtest[3,4,63]andFisher’sexacttestforanalyzingcarcinogenicitydata.Unlessotherwisenotedinthisdocument,allp-valuespresentedinthissectiononanimalcancerstudieswererecalculatedonmycomputerandaretheexactone-sidedp-valuesfortheFishertest(pFisher)and/ortheCochran-Armitagelineartrendtest(pTrend)whereappropriate.Incaseswherethedataispooledandthenumbersoftumorsarelarge,theapproximatep-valuebaseduponthenormaldistributionisusedforthetrendtesttoavoidexcessivecomputationtime;thesearenotedaspTrendA.Theapproximation(pTrendA)isgenerallyequivalenttotheexactp-value(pTrend)whentherearemorethan10animalswithtumors[64].


21

Toavoiddoinglargenumbersoftestsandover-analyzingthedata,mycommentswillgenerallyrelyupontheuseofthetrendtestwiththeresultsfromFisher’sexacttestservingasadescriptivediscussionofthefindings.ThisisinagreementwithSAPcomments[54]andisgenerallyacceptedintheevaluationofanimalcancerstudies.

Evenwiththehighdosesusedinthesestudies,itissometimesnecessarytouse“historicalcontrols”toevaluateagivenresponse.Historicalcontrolsaregenerallythehistoricalcollectionoftumorresponsesfromuntreatedcontrolgroupsfromstudiesinthesamelaboratorywithintwotothreeyearsofthestudybeingevaluated[30,34,59,65,66].Evaluationofthedatausingthehistoricalcontrolsshouldbedonerigorouslytocorrectlyevaluatetheresponsesseeninagivenstudy.Whereavalidhistoricalcontroldatasetwasavailable,Iusedthemeantumorresponseinthecontrolstocalculatetheprobabilityofobservingthetrendseeninthestudyoramoresignificanttrendifthetrueprobabilityofresponseisthehistoricalcontrolaverage;thisislabeledpHist.Inallcases,theguidelinesandliteraturesupporttheuseofthecontrolinthecurrentstudyasthemostappropriatecontrolgrouptouseunlessthereisaspecificneedtoaddresshistoricalresponses.Manyguidelines[30,33,34,67]suggesthistoricalcontrolsbeusedforevaluatingraretumorsandfindingsinassaysthatappeartobeunusual.Itisexplicitlynotedthatsignificantincreasesintumorsoverwhatisseenintheconcurrentcontrolshouldnotberejectedsimplybecausethetumorsareintherangeofthehistoricalcontrols[30].Norisitrecommendedtorejectsignificantincreasesintumorresponsesbecausethecontrolresponseisonthelowendofthehistoricalrange.Animalsarerandomlyassignedtocontrolandexposuregroupsandanylowresponseincontrolsislikelytoalsoreflectsimilarresponsepatternsintreatedanimals.ThisisinagreementwithSAPcomments[54]ontheEPAissuepaperonglyphosate[61]andwithallguidelinesforanalyzinganimalcarcinogenicitydata.

Thereare13animalcarcinogenicitystudiesinrats[68-80]andeightinmice[81-88].Onlytwostudies[71,77]appearinthepeer-reviewedliterature;theremainingstudiesarepartiallyavailablethroughseveralsources.Forthreeoftheratstudies[70,74,78]andtwomousestudies[83,86],technicalreportsfromtheperforminglaboratoryareavailablefromdocumentsprovidedbytheregistrant.Fortheremainingunpublishedstudies,datawasobtainedfromtheEPAreviewofglyphosate[61],theEuropeanFoodSafetyAuthorityreviewofglyphosate[89,90]andsupplementalmaterialfromareviewofthecarcinogenicityofglyphosatebyapanelofscientistsonbehalfofMonsanto[91].

Manyadditionalendpoints,otherthancancerincidenceandrelatedtoxicities,wereevaluatedinthesestudies;Iwillonlyprovidecommentsonthetumorincidencedataandrelateddatawhererelevanttothecancerfindings.

Itisunusualtohavemultiplecarcinogenicitystudiesinthesameexperimentalanimalmodelarisingfromdifferentlaboratories.Methodsforthecombinedanalysisofmultipleanimalcancerbioassaysarenotavailableinthescientificliterature.However,pooledanalyses,asconductedinepidemiology[92,93]areapplicableforcombininganimalcarcinogenicitystudies.Thebasicconceptistopoolalldatafromthesamesex/species/strainintoonestudyandanalyzeitappropriately.Thebasicstepsare:1)selectthestudiestobepooled;2)mergethedataforanalysis;3)estimatestudyspecific


22

effects;4)estimatepooledeffects;5)explainthedifferencesbetweenthepooledeffectsandtheindividualstudyeffects;6)doasensitivityanalysisifpossible.Thesestepswillbeusedtoanalyzepooleddatafromanimalcarcinogenicitystudieswherepoolingisdonebysex,species,strainanddurationofexposuretolimitheterogeneityacrosspooledstudies.IntheirrecommendationstotheEPAregardingEPA’sissuepaperonthecarcinogenicityofglyphosate[54],theFIFRAScienceAdvisorypanelstronglysupportedtheuseofapooledanalysistoaddressthequestionofconsistencycitingmycommentstotheEPA[94]. RatStudies

ReynaandGordon(1974)[76]exposedAlbinorats(probablySprague-Dawley)toammoniumsaltofglyphosate(13.85%purity)inatwo-yearchronicfeedingstudy.OnlyEPA[61]reportedonthisstudyandprovidednodetailsotherthantoreporttherewereapproximately70animalspergroupandtherewasinsufficientreportingonthehistopathologyfindings.Insufficientdetailisavailableonthisstudy.

Thisstudyisinadequateforuseindecidingoncausality.

Burnettetal.(1979)[70]exposedmaleandfemalealbinoratstoanaqueousmonosodiumsaltsolutionofglyphosatebyoralintubation(puritynotgiven).Therewere90animalspergroupanddoseswere0,3,10and30mg/kg/dayfor24months.EPA[61]reportedthatnohistopathologicalalterationswereobserved;noadditionalinformationwasavailableonthisstudy.ThisstudyhadseverelyreducedsensitivitytoobserveanycancerfindingsbecausethehighestdoseusedinthisstudyisverylowcomparedtotheMTDsintheotherratstudies.Thisstudydoesnotcontributetotheevaluationofcancercausationinlaboratoryanimalsandwillbeexcludedfromanyfurtherdiscussion.

Lankasetal.(1981)[74]exposedgroupsof50maleand50femaleSprague-Dawleyratstoglyphosate(98.7%purity)infeed(seeTable1fordoses)for26months.ThisstudyisnotinconcordancewithOECDguidelines(theywerenotavailableatthetimeofthisstudy),butasnotedbyEFSA[89],itwasingeneralaccordancewiththe1981OECDguidelines.InformationonthisstudywasavailablefromEPA[61],EFSA[89],Greimetal.[91],theoriginalstudyreportfromBio/dynamicsInc.[95]andmemosfromMonsantotoEPAprovidedbyMonsanto.

Therewerenosurvivaldifferencesinthisstudyandtherewasnoindicationthatthehighestdoseusedexceededthemaximum-tolerateddose.

Table1showsthestatisticallysignificanttrendintesticularinterstitialcelltumorsthatwasobserved(pTrend=0.009).Historicalcontrolswereprovidedinthestudyreportforfivestudieswithresponseratesof4/116,5/75,4/113,6/113and5/118forameanresponseof4.5%(24/535).ComparingthishistoricalcontrolmeantotheobservedresponseyieldspHist=0.006,showingthatthisresultissignificant,evenwhencomparingittothehistoricalcontroldataset.Lankasetal.(1981)arguedthatthetumorratesatsacrificewerenotstatisticallysignificantfromcontrolsuggestingthisfindingisnotrelatedtoglyphosate.However,byreducingthenumbersofanimalstoonlythoseat


23

terminalsacrifice,thepowertofindaneffectwassignificantlyreduced.Also,ifthetumorincreasestheanimal’schancesofdying,thensomeanimalswithtumorswilldieearly,whichcouldbiasresultsonlyseenatterminalsacrifice.Thistypeofanalysisissimplyneverdone;itappearstohavebeendevelopedforthiscasetodismisstheeffectsseeninthestudy.Lankasetal.(1981)alsosuggestedthecontrolresponsewaslowcomparedtothehistoricalrates,buttheconcurrentcontrolisalwaysthebestcontrolgrouptouseunlessitisclearlyflawed[33,34,59];inthiscase,therewasnoapparentproblemwiththecontrolsbecausetheprobabilityofseeing0/50ifthetruebackgroundresponseis4.5%isabout10%andthiscontrolgroupisnotsignificantlydifferentthanthehistoricalcontrols.EFSA[89]notedratesforinterstitialcellhyperplasia(apotentialprecursorfortheinterstitialcelltumors)andsawnodose-responsetrend(Table1).However,theseverylowrateswouldsuggestthatthetumorsarisinginthe10animalsthatdidgetinterstitialcelltumorsareindependentofamechanisminvolvinginterstitialcellhyperplasia.Thetumorresponseforinterstitialcelltumorswasnotmonotonic(tumorratesincreasingasdoseincreases),butwasstillwithinstatisticalvariation.TheEPASAPagrees,concludingthat“requiringvisualconfirmationofamonotonictrendinscatterplotsofdata…isknowntobeapoorwayofassessingtrend”[54].

AnincreaseinThyroidC-cellcarcinomas(Table1)wasobservedinfemalerats(pTrend=0.003)butcombiningadenomasandcarcinomaswasonlymarginallysignificant(pTrend=0.072).IndependentpathologistsbroughtinbyMonsantoarguedthesetumorswerenottreatmentrelated.Theauthorsprovidedhistoricalcontroldataforbothcarcinomasandcarcinomascombinedwithadenomasfromninecontrolgroupswithmeanresponsesof4/453=0.9%forcarcinomasand46/453=10.2%forthecombinedtumors.Thesignificanceofbothresultswasunchangedusingthehistoricalcontroldata.

Theauthorsalsomentionedthattheincidenceoflymphocytichyperplasiainthethymusandlymphnodeswereslightlyelevatedabovecontrols(pTrend=0.143).Themiddledosegroupwassignificantlydifferentfromcontrols(pFisher=0.018).

Thisstudyalsohadastatisticallysignificantincreaseinpancreaticisletcelltumorsinthelowestdose(pFisher=0.028)inmales(Table1),butnotanyoftheotherdoses;thetrendtestwasnotsignificant(pTrend=0.312).

ThehighestdoseusedinthisstudyinSprague-DawleyratsisfarbelowtheMTD.EventhoughEFSA[89]notedthatthisstudywasingeneralaccordancewiththe1981OECDguidelines,theydismisseditfornotmeetingcurrentguidelinesduetothelow-dosesused.EPA[61]alsoexcludedthisstudyfromconsideration.However,thestudysawanincreaseintesticulartumorsinmalesandThyroidC-cellcarcinomasinfemalesthatshouldbecarefullyevaluatedindeterminingcausality.Also,thisisthestudywiththelongestexposure(26months)andprovidesuniqueinformationtotheoverallevaluation.

AdditionaltumorsseentohavesignificantincreasesinotherstudiesusingSprague-DawleyRatsarealsoincludedinTable1.


24

Inconclusion,thisstudyshowspositiveresultfortestesinterstitialcelltumorsandhepatocellularadenomasinmaleSprague-Dawleyratsandapositiveresponseforthyroidc-cellcarcinomasinfemaleSprague-Dawleyratsandwillbeincludedintheoverallevaluationofcausation.

StoutandRuecker(1990)[78]exposedgroupsof50maleand50femaleSprague-Dawleyratstoglyphosate(98.7%purity)infeed(seeTable2fordoses)for24months.ThisstudywasdoneunderOECDguidelines.


Pancreaticisletcelltumorswereincreasedinalldosegroupsrelativetothecontrolsinmaleratsandstatisticallysignificantforthelowest(pFisher=0.015)andhighest(pFisher=0.032)dosegroups(Table2).However,theseratesincludethe10animalsthatweresacrificedatoneyear.Duetotheshortdurationofexposure,theratsterminatedatoneyearwerelikelynotatriskofdevelopingthistumor;itisveryunusualtoincludetheseanimalsinthefinaltumorcounts(EPA[61]alsoexcludedtheseanimals).Inthepathologytablesforthisstudy,therewerenotumorsinanyofthe10animalsattheinterimsacrifice.Removingthese10animalsdoesnotalterthep-valuesfortrendor

Table1:TumorsofinterestinmaleandfemaleSprague-Dawleyratsthe26-monthfeedingstudyofLankas(1981)[74]

Tumor Sex Doses(mg/kg/day) p-values

Male 0 3.05 10.30 31.49Female 0 3.37 11.22 34.02

Testicularinterstitialcelltumors

Male 0/50 3/50 1/50 6/50** PTrend=0.009PHist=0.006

Interstitialcellhyperplasia Male 1/50 1/50 1/50 0/50 PTrend=0.830ThyroidC-cellCarcinomas Female 1/47 0/49 2/50 6/47 PTrend=0.003

PHist=<0.001ThyroidC-cellAdenomasandCarcinomas

Female 6/47 3/49 8/50 9/47 PTrend=0.072PHist=0.072

PancreasIsletCellTumors Male 0/50 5/50* 2/50 3/50 PTrend=0.312lymphocytichyperplasia,thymusandlymphnodes

Female 27/50 35/50 38/50* 35/50 PTrend=0.143

ThyroidC-cellAdenomasandCarcinomas

Male 1/47 2/49 4/49 4/49 PTrend=0.122

ThyroidFollicular-cellAdenoma

Male 5/47 1/49 2/49 2/49 PTrend=0.748

LiverNeoplasticNodule Male 3/50 5/50 1/50 3/10 PTrend=0.630KidneyAdenoma Male 1/50 5/50 0/50 0/50 PTrend=0.979*-pFisher<0.05,**-pFisher<0.01


25

Fisher’sexacttest.Historicalcontroldataforthistumorinthislaboratorywasreportedas23/432or5.3%[96]andatrendcomparisonagainstthiscontrolratewasnotsignificant(phist=0.15).Thelackofatrendisdrivenbytheupanddownnatureoftheresponse.Assumingthehistoricalrateof5.3%iscorrect,thechancesofseeingeightormoretumorsin47animalsis0.003.Similarly,forthemid-andhigh-doses,thisprobabilityis0.124and0.014,respectively.Femalesdidnotshowanincreaseinthistumor.Theauthorsprovidedatablewiththecombinedresultsforpancreaticislet-celladenomasandcarcinomasfromthisstudywiththetumorcountsfromtheLankasetal.(1981)[74]studyarguingtheresultsdonotshowadose-relatedincrease.Animalsstudiedfor26monthsversus24monthscanhaveverydifferentresponsestothesamechemicalandverydifferentcontrolincidence.

Inmalerats,therewasastatisticallysignificanttrend(pTrend=0.015)afterremovalofinterim-sacrificedanimalsforhepatocellularadenomasbutasignificantincreaseforadenomasandcarcinomascombined(pTrend=0.05,Table2)andnotinfemales(notshown).Livercarcinomasaregenerallyalsoprovidedinaseparateanalysis,butthesedatawerenotprovidedbytheauthors(thedatawouldsuggestthehepatocellularcarcinomaswouldhaveanegativetrend).

TherewasalsoasignificantincreaseinthyroidC-celladenomasinthefemalerats(pTrend=0.049)andamarginalincrease6inadenomasandcarcinomascombined(pTrend=0.052)regardlessofwhetherinterimsacrificedanimalsareincluded(Table2).Inmales,thetrendforadenomaswaspTrend=0.084andforadenomasandcarcinomaswaspTrend=0.091.Adenomaswereseeninmaleratsattheinterimsacrificedemonstratingthatmaleratsattheinterimsacrificewereatriskforthistumor.Iftheseanimalsareaddedbackintotheanalysis,thetrendtestinmaleshaspTrend=0.063foradenomasandpTrend=0.068foradenomasandcarcinomascombined.

SeveralothertumorsdemonstratingsignificantfindingsinotherstudiesofSprague-DawleyratsareincludedinTable2anddonotshowsignificanteffects.

Inconclusion,thefindingofanincreasedincidenceofpancreaticislet-celltumorsinthisstudycannoteasilyberuledoutasachancefinding.Findingsofsignificantincreasesinliveradenomasinmaleratswithnoincreasesincarcinomascouldbeduetochance.Thefindingsofsignificantincreasesinthyroidc-celltumorsinmalesandfemalesshouldbecomparedwithotherstudies.Thisstudywillbeincludedintheoverallevaluationofcausation.

6Instatistics,itiscommontorefertop-valuesintherangeof0.10>p-value>0.05asmarginalwhenthetargetp-valueis≤0.05;thisisdonetoavoidmissingtrendsindatareflectedbyalmostsignificantfindings


26

Table2:TumorsofinterestinmaleandfemaleSprague-Dawleyratsfromthe24-monthfeedingstudyofStoutandRuecker(1990)[78]


Male 0 89 362 940Female 0 113 457 1183

PancreasIsletCellTumors(withinterimsacrifice)

Male 1/58 8/57* 5/60 7/59* PTrend=0.147PHist=0.140

PancreasIsletCellTumors(withoutinterimsacrifice)

Male 1/48 8/47* 5/50 7/49* PTrend=0.147PHist=0.150

Hepatocellularadenomas(withoutinterimsacrifice)

Male 3/50 2/50 3/50 8/50 PTrend=0.015

HepatocellularAdenomasandCarcinomas(withoutinterimsacrifice)

Male 6/50 4/50 4/50 10/50 PTrend=0.050

ThyroidC-CellAdenomas(withinterimsacrifice)

Female 2/60 2/60 6/60 6/60 PTrend=0.050

ThyroidC-CellAdenomas(withoutinterimsacrifice)

Female 2/50 2/50 6/50 6/50 PTrend=0.049

ThyroidC-CellAdenomasandCarcinomas(withinterimsacrifice)

Female 2/60 2/60 7/60 6/60 PTrend=0.053

ThyroidC-CellAdenomasandCarcinomas(withoutinterimsacrifice)

Female 2/50 2/50 7/50 6/50 PTrend=0.052

ThyroidC-CellAdenomas(withinterimsacrifice)

Male 2/60 4/60 8/60 7/60 PTrend=0.063

ThyroidC-CellAdenomas(withoutinterimsacrifice)

Male 0/50 4/50 8/50** 5/50* PTrend=0.084

ThyroidC-CellAdenomasandCarcinomas(withinterimsacrifice)

Male 2/60 6/60 8/60* 8/60* PTrend=0.068

ThyroidC-CellAdenomasandCarcinomas(withoutinterimsacrifice)

Male 0/50 6/50* 8/50** 6/50* PTrend=0.091

TestisInterstitialCellTumors Male 2/50 0/50 3/50 2/50 PTrend=0.296KidneyAdenomas Males 0/50 2/50 0/50 0/50 PTrend=0.813ThyroidFollicularAdenoma/Carcinoma

Males 2/50 1/48 3/48 3/50 PTrend=0.225

*-pFisher<0.05,**-pFisher<0.01


27

Atkinsonetal.(1993)[68]conductedacombinedchronictoxicity/carcinogenicitystudyofglyphosate(98.9%pure).Theyused50Sprague-DawleyratsineachgroupforbothsexeswithdietaryexposuresgiveninTable3.Anadditional35rats/sex/dosewereincludedforinterimsacrifices.


Table3:TumorsofinterestinmaleandfemaleSprague-Dawleyratsfromthe24-monthfeedingstudyofAtkinsonetal.(1993)[68]


Male 0 11 112 320 1147Female 0 12 109 347 1134

ThyroidFollicularAdenomasandCarcinomas

Male 0/50 0/21 0/17 2/21 2/49 PTrend=0.099

ThyroidFollicularAdenomasandCarcinomas(addingterminalsacrificeanimalstodenominator)

Male 0/50 0/50 0/50 2/50 2/49 PTrend=0.034


Female 8/50 1/27 1/29 1/29 7/49 PTrend=0.197


Male 9/50 1/21 1/17 2/21 9/49 PTrend=0.183

TestesInterstitialCellTumors

Male 3/50 1/25 0/19 0/21 2/50 PTrend=0.580

KidneyAdenomas Males 1/50 0/50 0/50 0/50 0/50 pTrend=1HepatocellularAdenomas Males 2/50 1/50 1/50 2/50 3/50 PTrend=0.155PancreasIslet-CellAdenoma Male 0/50 0/50 0/50 0/50 1/50 PTrend=0.200*-pFisher<0.05,**-pFisher<0.01

Theauthorsreportednosignificanteffects,asdoEPA[61]andEFSA[89].Thestudydidnotdodetailedhistopathologicalexaminationonallanimalsinallgroupsforeverytumortype,butdidexamineallcontrolandhighdoseanimals,allanimalsthatdiedbeforestudyterminationandanimalsshowingmacroscopictumorsatstudytermination;liver,kidneyandlungswereexaminedforallanimals.Thisseverelyweakensthestudyforaddressingdose-responsetrends.However,inreviewingthepathologytablesprovidedinGreimetal.(2015)[91],thyroidfollicularadenomasandcarcinomaswerefoundtobemarginallysignificant(pTrend=0.099)bythetrendtest.Ifthethreemiddleexposuregroupshadseennoothertumorsandthedenominatorsweretheentire50animalsonstudy,thetrendanalysisbecomessignificant(pTrend=0.034).

Withoutexaminationoftheanimalsfreeofgrosstumorsatterminalsacrifice,thefindingsfromthisstudywillbegivenlessweightintheoverallevaluationofcausation.

Brammer(2001)[69]conductedatwo-yearcarcinogenicitystudyinWistarratsinwhichgroupsof52animalswereexposedtoglyphosate(97.6%pure)atdosesprovidedin


28

Table4.Anadditional12animalsweresacrificedatone-year.

AsignificantpositivetrendinsurvivalwasnotedbytheEPA(p=0.03),howeverthistrendwasnotaccomplishedusingaKaplan-Meirtest[97](theappropriatetest),butsimplyatestrelatingtothepercentsurvivingtoterminalsacrifice.Therewasnoindicationthatthehighestdoseusedexceededthemaximum-tolerateddose.

EPA[61],butnotEFSA[89],notedtherewasastatisticallysignificanttrendofhepatocellularadenomasinmaleratswiththehighestdosealsobeingstatisticallysignificantfromthecontrol.TrendanalysisgivespTrend=0.008andtheFisher’sexacttestcomparisonofhighdosetocontrolispFisher=0.027.EPAdismissedthisfindingaspotentiallyduetoaslightdifferenceinthenumberofanimalsattheterminalsacrificeinthisstudyversuscontrols.However,noformalstatisticalevaluationofsurvivalisprovidedanditcannotbeassumedfromthesenumbersthatsurvivalwassignificantlyimpactedintheseanimals.Greimetal.(2015)[91]usedslightlydifferentnumbersforthistumorbecausethreeanimals(oneinthecontrolgroup,oneinthelow-dosegroupandoneinthemid-dosegroup)intheinterimsacrificegroupdiedbeforetheirsacrificetimeand,fromthepathologytablesprovidedintheirpaper,thesecouldnotbeseparatedfromothers.ThesenumbershavebeenincludedinTable4,butitdoesnotchangethesignificanceofthefindings.Greimetal.(2015)[91]dismissedthesefindings,partlybecauseofthesamesurvivalargumentusedbytheEPAandpartlybecausetheyhadahistoricalcontroldatasetwheretherangeofhistoricalresponsewasfrom0-11.5%;theydidnotprovidethemeanresponseortheindividualtumorresponsesforthesehistoricalcontrols.Asmentionedearlier,dismissingresultsbecausetheyareintherangeofthehistoricalcontrolsisanunacceptablemethodforusinghistoricalcontrolstoevaluateastudy,andinthiscase,thereisnoreasontoquestiontheconcurrentcontrols.

Table4:TumorsofinterestinmaleandfemaleWistarratsfromthe24-monthfeedingstudyofBrammer(2001)[69]

Tumor Sex Doses(mg/kg/day) p-valuesMale 0 121 361 1214Female 0 145 437 1498

HepatocellularAdenoma Male 0/52 2/52 0/52 5/52* PTrend=0.008HepatocellularAdenoma(fromGreimetal.,2015[91])

Male 0/53 2/53 0/53 5/52* PTrend=0.008PHist=0.006

MammaryGlandAdenomasandAdenocarcinomas

Female 3/51 2/51 0/51 2/51 PTrend=0.575

SkinKeratocanthoma Male 1/51 0/51 1/51 1/51 PTrend=0.392*-pFisher<0.05,**-pFisher<0.01

Iobtainedhistoricalcontroldatafrom16controlgroupsinWistarratsfromCharlesRiverLaboratoriesfortheyears2003to2011[98].AlthoughtheseareoutsideoftheoptimaltimerangefortheanimalsusedintheBrammer(2001)study,theycanserveasanillustrationofwhyusingarangecanbemisleading.Therewere52liveradenomas


29

seenin1217controlanimalsforameanresponseof4.27%witharangeof0%to17.5%(individualstudyfindingsof6/100,0/60,1/60,1/50,1/80,14/112,1/65,0/60,21/120,0/50,1/50,2/60,0/50,1/100,1/150,2/50;13studieswith≤2%response).Assumingtheunderlyingprobabilityofhavingatumorincontrolsis4.27%,pHist=0.006(Table4).Thus,eventhoughtheresponsesseeninBrammer(2001)areintherangeofthehistoricalcontrols,thetrendishighlysignificantwhenhistoricalcontrolsareusedappropriately.Greimetal.(2015)alsomentionedfindingsofincreasedtoxicityatthehighdoseforwhichtheyprovidednumbersforonlyhepatocytefatvacuolationandhepatitis;noneofthesefindingswerestatisticallysignificantbyanytest.

Inconclusion,thisstudyshowsapositiveresultforhepatocellularadenomasinmaleWistarratsandwillbeincludedintheoverallevaluationofcausation.

PavkovandWyand(1987)[75]exposedSprague-Dawleyratstoglyphosatetrimesiumsalt(sulfosate,56.2%pure)infeedfortwoyears.Eightyanimals/sexweretestedinthecontrol,low-doseandmid-dosegroups,and90/sexweretestedinthehighdosegroup.Dosesof0,4.2,21.2and41.8mg/kg/daywereusedinmalesand0,5.4,27,and55.7mg/kg/dayinfemales.ThisstudyshowednosignificantfindingsaccordingtoEPA[61].Nodetailsweregivenbeyondthatsimplestatementandnoothersreportedonthisstudy.ThedosesinthisstudyarefarbelowtheMTDsothisstudywouldhavereducedsensitivitytodetectaneffectifoneexisted.ThisstudyalsousedadifferentchemicalthantheotherSprague-Dawleyratstudiesandisnotcomparableonthatbasis.

Thisstudyisnotacceptableforuseintheevaluationofcausalityduetothelackofdetailsaboutthestudy.

Suresh,(1996)[79]exposedWistarratstoglyphosate(96.8%pure)infeedfortwoyears.Fiftyanimals/sexweretestedinfourexposuregroupsshowninTable5.


EPA[61]concludedtherewerenotumorsincreasedduetoglyphosateexposureinthisstudyandEFSA[89]concludedthat,“[n]oneofthesignificantmicroscopicchanges,increasedanddecreasedincidences(inliver,spleen,lymphnodes,adrenals,thymus,gonads,uterus,mammarygland)observedhaveshowndoserelationship,henceappearedtobeincidentalandnotrelatedtothetreatmentwiththetestcompound.”(page491).Greimetal.(2015)[91]provideddataonhepatocellularadenomasandcarcinomasinbothsexesbutnoneoftheseshowedsignificanttrendsorpairwisetests(Table5).However,therewasanotherstudywithastrongsignificanttrendinhepatocellularadenomasinWistarrats[69]sothesearealsoincludedinTable5forcomparison.NoothertumorswerementionedbyanyothergroupandanexaminationofthegroupedpathologytablesprovidedbyGreimetal.(2015)showanincreaseinmammaryglandadenomasatthemid-dose(pFisher=0.017)butnosignificanttrend.However,therewasanotherstudywithastrongsignificanttrendinmammaryglandadenomasandadenocarcinomascombinedinWistarrats[80]sothesearealsoincludedinTable5forcomparison.LiketheAtkinsonetal.(1993)[68]study,Suresh(1996)didnotdofullpathologyonalloftheanimalsintheinterimexposuregroupsmaking


30

interpretationofthisstudyproblematic.

Thisstudywillbeincludedintheoverallevaluationofcausation.

Table5:TumorsofinterestinmaleandfemaleWistarratsfromthe24-monthfeedingstudyofSuresh(1996)[79]

Tumor Sex Doses(mg/kg/day) p-valuesMale 0 6.3 59.4 595.2Female 0 8.6 88.5 886

MammaryGlandAdenomaandCarcinoma

Female 5/40 3/28 8/33 2/48 PTrend=0.970

HepatocellularAdenoma Male 24/50 22/50 10/50 21/50 PTrend=0.374SkinKeratocanthoma Male 0/50 0/50 0/50 0/50 PTrend=1*-pFisher<0.05,**-pFisher<0.01

Enemoto(1997)[72]exposedSprague-Dawleyratstoglyphosate(95.7%pure)infeedfortwoyears.Fiftyanimals/sexweretestedinfourexposuregroups(seeTable6).Inaddition,10animalsperexposuregroupwereexposedfor1yearandanother10for18monthsatwhichpointtheyweresacrificedandexamined.Theseinterimsacrificeanimals(1yearand18months)areincludedintheanalysisiftumorswereseeninthesegroups.

Therewerenosurvivaldifferencesinthisstudyandtherewasnoindicationthatthehighestdoseexceededthemaximum-tolerateddose.

EPAandEFSAbothfoundnosignificantchangesintumorsinanygroup.Greimetal.(2015)againprovidetablesforanumberoftumors,noneofwhichshowsignificanteffectsexceptfortheincidenceofkidneyadenomasinmalerats(pTrend=0.004,Table6).ExaminingthepathologytablesprovidedinGreimetal.(2015)revealsnoadditionaltumorsshowinganincreaseintumorincidencewithdose.Adifferentstudy[74]inSprague-Dawleyratsdemonstratedastrongsignificanttrendinmammaryglandadenomas,thyroidC-cellcarcinomas,skinKeratocanthomasandtesticularinterstitialcelltumorssothesearealsoincludedinTable6forcomparison.

Thisstudyshowedasignificantincreaseinkidneyadenomasandwillbeincludedintheoverallevaluationofcausation.


31

Woodetal.(2009)[80]exposedWistarratstoglyphosate(94.7%to97.6%pure)infeedfortwoyears.Fifty-oneanimals/sexweretestedinfourexposuregroupsatdosesshowninTable7.

Nosurvivaldifferenceswereseeninthisstudy.

EFSA[89]foundnodose-relatedtumorincreaseswhileEPA[61]notedanincreaseinmammaryglandadenomasandadenocarcinomascombinedwithpTrend=0.062foradenomas,pTrend=0.042foradenocarcinomasandpTrend=0.007forthecombinedtumors(Table7).EPAconcludedtherewasnoprogressionfromadenomatoadenocarcinomaandarguedtheincreasewasnotglyphosaterelated.Thisconclusioniscontradictedbythefactthat6animalsincontrolandthelowerdosegroupsgotcarcinomaswithnoadenomasinanyoftheanimalsinthesegroups.Itseemslikelythat,inthiscase,mammaryglandadenocarcinomascanarisewithoutthepresenceofanyadenomas.Greimetal(2015)[91]alsonotedanincreaseinskinkeratoacanthomainmales(pTrend=0.030).Reviewofthepathologytablesidentifiednoothertumorswithincreasedtumorratesasafunctionofdose.TherewasanotherstudywithastrongsignificanttrendinhepatocellularadenomasinWistarrats[69]sothistumorisalsoincludedinTable7forcomparison.

Thisstudyshowedanincreaseinmammarytumorsinfemalesandskinkeratoacanthomasinmalesandwillbeusedintheevaluationofcausality.

Table6:TumorsofinterestinmaleandfemaleSprague-Dawleyratsfromthe24-monthfeedingstudyofEnemoto(1997)[72]


Male 0 104 354 1127Female 0 115 393 1247

MammaryGlandAdenoma Female 23/50 27/50 24/50 30/50 PTrend=0.106KidneyAdenoma Male 0/50 0/50 0/50 4/50 PTrend=0.004ThyroidC-cellAdenomas/Carcinomas

Female 4/60 7/60 8/60 4/60 PTrend=0.692

ThyroidC-cellAdenomas/Carcinomas

Male 8/70 10/70 6/70 7/70 PTrend=0.697

ThyroidFollicular-cellAdenomas/Carcinomas

Male 4/70 2/70 1/70 0/70 PTrend=0.990

TestesInterstitialCellTumors

Male 3/49 2/50 0/50 2/50 PTrend=0.594

HepatocellularAdenomas Male 1/60 0/60 2/60 1/60 PTrend=0.371SkinKeratocanthoma Male 3/50 3/50 0/50 6/50 PTrend=0.065PancreasIslet-CellAdenoma Male 4/50 1/50 2/50 1/50 PTrend=0.844*-pFisher<0.05,**-pFisher<0.01


32

Table7:TumorsofinterestinmaleandfemaleWistarratsfromthe24-monthfeedingstudyofWoodetal.(2009)[80]

Tumor Sex Doses(mg/kg/day) p-valuesMale 0 85.5 285.2 1077.4Female 0 104.5 348.6 1381.9

MammaryGlandAdenomas Female 0/51 0/51 0/51 2/51 PTrend=0.062MammaryGlandAdenocarcinomas

Female 2/51 3/51 1/51 6/51 PTrend=0.042

MammaryGlandAdenomasandAdenocarcinomas

Female 2/51 3/51 1/51 8/51* PTrend=0.007

SkinKeratocanthoma Male 2/51 3/51 0/51 6/51 PTrend=0.030

HepatocellularAdenoma Male 0/51 2/51 1/51 1/51 PTrend=0.418

*-pFisher<0.05,**-pFisher<0.01

Excel(1997)[73]exposedSprague-Dawleyratstoglyphosate(puritynotgiven)infeedfortwoyears.Fifty-oneanimals/sexweretestedinfourexposuregroupsatdosesof0,150,780and1290mg/kg/dayinmalesand0,210,1060and1740mg/kg/dayinfemales.EPA[61],EFSA[89]andGreimetal.(2015)[91]hadconcernswiththequalityofthisstudy,thecharacterizationofthechemicalbeingusedandwithtumorratesinthisstrainofanimalsbeingtoolow.TheSupplementalMaterialfromGreimetal.(2015)onthisstudyshowsnosignificantincreaseinanytumorandvirtuallyallanimalshavingnotumorsincontrolsandtreatedanimals.

ThisstudyisinadequateforuseindecidingoncausalityforthesamereasonsgivenbytheEPA,EFSAandGreimetal.(2015).

Chruscielska,K.(2000)[71]exposedWistarratstoglyphosateasa13.8%solution(puritynotgiven)indrinkingwaterfortwoyears.AccordingtoGreimetal.(2015)[91],thisappearstobetheglyphosateformulationPerzocyd.Eighty-fiveanimals/sexweretestedinfourexposuregroups.Theauthorslistedthedosesascontrol,300mg/L,900mg/Land2700mg/Lindrinkingwater.Greimetal.(2015)[91]estimatedtheintakeofglyphosatetobe0,1.9,5.7and17mg/kg/dayforfemalesand0,2.2,6.5,and19mg/kg/dayinmales.Therewasaslightincreaseinmalignantadenomasofthepituitaryglandandanoppositedecreaseinpituitaryadenomassuggestingnoeffectorpotentiallyapromotionaleffectinwhichadenomasarepromotedtocarcinomasbyglyphosate.Nootherincreasedtumorresponseswerereportedinthemanuscript.Becauseofthelowexposures,thisstudyisaninadequatechallengetotheanimals(thehighestdoseisfarbelowtheMTD).Thereportingofthisstudyisverylimitedandittheoverallqualityoftheworkcannotbeevaluated.

Thisstudyisinadequateforuseindecidingoncausality.

Seralini,G.E.,etal.(2014)[77]exposedSprague-DawleyratstotheglyphosateformulationRoundupindrinkingwaterfortwoyearsaspartofabroaderexperimenton


33

Roundup-ReadyCorn.Tenanimals/sexweretestedinfourexposuregroupsatdosesof0,0.00005,400and22500mg/Linfemales.Theauthorsreportedanincreaseintheincidenceofmammaryglandtumors(mainlyfibroadenomasandadenocarcinomas)infemaleratswithincidencesof5/10forcontroland9/10,10/10,9/10(pFisher=0.016)inthelow-,mid-andhigh-dosesgroupsrespectively.Itisdifficulttoassessthequalityofthisstudyduetolimitedreportingonthehistopathologicaldescriptionsofthetumorsandtheverysmallsamplesize.

Thisstudywillnotbeusedintheevaluationofcausality.

JointAnalysis-Rats

Table8summarizesthesignificanceforalltumorsofinterestinrats.

Brammer(2001)[69]sawasignificantincreaseinhepatocellularadenomasinmaleWistarratswithincreasingdose(pTrend=0.008,Table4).TheothertwoacceptablestudiesinWistarrats(Woodetal.(2009)[80]andSuresh(1996)[79]didnotseesignificantincreases(Tables5and7).Onthebasisofstatisticalsignificance,thesestudiesareinconsistent.Torejectthesefindingsbasedupononly1/3beingpositiveisthesameasrejectingacoinasbeingfairif,inthreeflipsofthecoin,theresultisoneheadandtwotails;itsimplyisnotpossibleandthereisabetterwaytoaddressthesefindings.Givendifferentdosesanddifferentsamplesizes,weneedtoformallytestforconsistencyinthesestudies.Suresh(1996)saw48%responseforhepatocellularadenomasincontrolswhereastheothertwostudiessawnotumorsinthecontrolanimals.Thus,althoughallthreestudiesareinWistarrats,Suresh(1996)hasasignificantlydifferentcontrolresponsefromtheothertwo.Suresh(1996)didnotgiveasubstrainfortheWistarratsused,butBrammer(2001)andWoodetal.(2009)useddifferentsubstrains.Allthreestudiesuseddifferentdietsandwereconductedindifferentfacilities.Thus,thereisnoobviousexplanationforthedramaticallydifferentratesinSuresh(1996).Itisknownthatthesamestrainofratsfromdifferentlaboratoriescanhavemarkedlydifferentcontroltumorresponses.Becausetheyhavesimilarcontrolresponse,Brammer(2001)andWoodetal.(2009)canbepooledintoasinglestudytoaskthequestion“DoesthesignificanttrendforBrammer(2001)disappearwhenitispooledwiththenegativestudyofWoodetal.(2009)?”TheanalysisofthepooledstudiesyieldspTrend=0.013supportingtheconclusionthatglyphosatecauseshepatocellularadenomasinWistarratswithsimilarbackgroundresponses.

Woodetal.(2009)[80]sawasignificantincreaseinmammaryglandadenomasandadenocarcinomas(pTrend=0.007,Table7)infemalesthatwasnotseenintheothertwostudies(Tables4and6).ThebackgroundratesinthesestudiesdifferonlyslightlyandapooledanalysisofallthreestudiesyieldspTrendA=0.459,suggestingthatcombiningthedataeliminatesthedose-responsetrendseeninWoodetal.(2009).However,iftheWistarratsusedinSuresh(1996)differedintheirresponseforhepatocellularadenomas,theymaydifferforthistumoraswell.CombiningonlyWoodetal.(2009)withBrammer(2001)resultsinpTrend=0.037.GiventhemixedresultsfromthepoolingforthistumorIconcludethereislimitedsupportforthenotionthatglyphosatecancausemammaryglandadenomasandadenocarcinomasinWistarrats.


34

Woodetal.(2009)[80]sawasignificantincreaseinskinkeratocanthomas(pTrend=0.030,Table7)inmalesthatwasnotseenintheothertwostudies(Tables4and6).ThebackgroundratesinthesestudiesdifferonlyslightlyandapooledanalysisofallthreestudiesyieldspTrendA=0.010,suggestingthatcombiningthedatadoesnoteliminatethedose-responsetrendseeninWoodetal.(2009).CombiningonlyWoodetal.(2009)withBrammer(2001)resultsinpTrend=0.053.GiventheresultsfromthepoolingforthistumorIconcludethereissupportforthenotionthatglyphosatecancauseskinkeratocanthomasinWistarrats.

InSprague-Dawleyrats,therewerefourstudiesthatwereacceptableforinclusionintheevaluationofcausalitywithone[74]yieldingstrongpositiveresponsesforthyroidC-cellcarcinomasinfemalesandtesticularinterstitialtumorsandhepatocellularadenomasinmalesandanother[72]yieldingastrongresultforkidneyadenomasinmales.Lankas(1981)[74]sawasignificantincreaseinthyroidC-cellcarcinomasinfemaleratsexposedtoglyphosate(pTrend=0.003,Table1)andamarginalincreaseinC-celladenomasandcarcinomascombined(pTrend=0.072,phist=0.072,Table1;oneoftheotherthreestudiesalsosawmarginalresultsforthyroidC-celladenomasandcarcinomasinfemales(Table2).ApooledanalysisusingallfourstudiesyieldspTrendA=0.390.ThispooledanalysisdoesnotsupporttheresultsseeninLankas(1981).However,theLankas(1981)studywasfor26monthsandtheotherthreewerefor24months;theC-cellcarcinomascouldbearesultofthelongerexposureperiodeventhoughthedoseissubstantiallylowerinthisstudycomparedtotheothertwo.Fromthesedata,IconcludethattheevidenceisweakthatglyphosatecausesthyroidC-celltumorsinfemaleSprague-Dawleyrats.

ThyroidC-celladenomasandcarcinomascombined,inmales,showmarginallysignificantdose-responsetrendsinStoutandRuecker(1990,Table2)butnotintheremainingthreestudies.PoolingallfourstudiesyieldsasignificanttrendofpTrendA=0.041.Fromthesedata,IconcludethatthereisevidenceisthatglyphosatecausesthyroidC-celltumorsinmaleSprague-Dawleyrats.

Thyroidfollicular-celladenomasandcarcinomascombined,inmales,showasignificantdose-responsetrendinAtkinsonetal.(1993,Table3)butnotintheremainingthreestudies;.PoolingallfourstudiesyieldsnosignificanttrendwithpTrendA=0.618.Fromthesedata,Iconcludethatthereisnoevidencethatglyphosatecausesthyroidfollicular-celltumorsinmaleSprague-Dawleyrats.

Hepatocellularadenomas,inmales,showasignificantdose-responsetrendinStoutandRuecker(1990,Table2)butnotintheremainingthreestudies.PoolingallfourstudiesyieldsamarginallysignificanttrendwithpTrend=0.073.Fromthesedata,Iconcludethatthereislimitedevidencethatglyphosatecausesthyroidfollicular-celltumorsinmaleSprague-Dawleyrats.


35

AnothersignificanttrendseeninSprague-DawleyratsisthefindingoftestesinterstitialcelltumorsfromLankas(1981)[74](PTrend=0.009,Table1);theotherthreestudieswerenegativeforthistumor(Tables2,3and6).CombiningtheotherthreestudieswiththatofLankas(1981)fortestesinterstitialtumorsresultsinap-valuefortrendthatisclearlynon-significant(pTrendA=0.608).However,asnotedabove,theLankas(1981)studywasfor26monthsandtheothertwowerefor24months;thetumorscouldbearesultofthelongerexposureperiodeventhoughthedoseissubstantiallylowerinthisstudycomparedtoStoutandRuecker(1990),Atkinsonetal.(1993)andEnemoto(1997).

ThefinaltumorinSprague-Dawleyratsshowingastrongsignificanttrendiskidney

Table8:Summaryofsignificancetestsfor5tumorsfrom7studiesinRats

Study Strain NeoplasmHepato-cellular

Adenomas(males)

MammaryGlandTumors(females)

SkinKerato-

canthoma(males)

ThyroidC-CellTumors(females)

ThyroidC-CellTumors(males)

ThyroidFollicular

CellTumors(males)

TestisInter-stitialCell

Tumors(male)

KidneyAdenomas(males)

Brammer(2001)[69]

Wistar +++1 -

Wood(2009)[80]

- +++ ++

Suresh(1996)[79]

- -

PooledWistarRats ++2 ++2 +++ Lankas(1981)[74]

SpragueDawley

-3 + - - +++ -

Enemoto(1997)[72]

- - - - - +++

Atkinsonetal.

(1993)[68]

- - - ++ - -

Stoutand

Ruecker(1990)

++ + + - - -

PooledSprague-DawleyRats

+ - ++ - - ++4

1entriesarepTrend/pHistwithvalues:–p>0.1,+0.1≥p>0.05,++0.05≥p>0.01,+++p≤0.01;2poolingresultsfromBrammer(2001)andWood(2009)only;3liverneoplasticnodules;4excludingLankas(1981)


36

adenomasinmalesfromthestudybyEnemoto(1997)[72](PTrend=0.004,Table6).ThekidneytumordataisnotsignificantforthestudiesbyLankas(1981)[74](Table1),Atkinsonetal.(1993)[99](Table3)andStoutandRuecker(1990)[78](Table2).PoolingtheEnemoto(1997)studywiththatofLankas(1981)[74],StoutandRuecker(1990)andAtkinsonetal.(1993)yieldspTrendA=0.201.Removingthe26-monthstudybyLankas(1981)[74]yieldsap-valueforthethreecombined24-monthstudiesofpTrend=0.031;thus,theassociationbetweenglyphosateandkidneyadenomasinmaleSprague-Dawleyratsissupportedbythesedata,evenwiththedifficultyassociatedwithinterpretingtheresultsinthelow-andmid-dosesintheAtkinsonetal.(1993)study.ThereisevidencetosupportanincreaseinkidneytumorsinmaleSprague-Dawleyratsexposedtoglyphosate.

Insummary,thereisevidencethatglyphosatecauseshepatocellularadenomasandskinkeratocanthomasinmaleWistarrats,mammaryglandadenomasandadenocarcinomasinfemaleWistarratsandkidneyadenomasandthyroidC-celladenomasandcarcinomasinmaleSprague-Dawleyrats.ThereislimitedevidenceglyphosatecauseshepatocellularadenomasinmaleSprague-Dawleyrats.

MouseStudies

ReynaandGordon(1974)[86]exposedSwissWhitemicetoglyphosate(>97%purity)infeedfor16monthsinmalesand18monthsinfemales.Fiftyanimals/group/sexweretestedinthreeexposuregroups;control,17mg/kgand50mg/kg.Only10animalspergroupwereexaminedforhistopathologicalchanges.

TherewasnoimpactonsurvivalofadministrationofglyphosateandnoindicationthatthehighdoseexceededtheMTD.

Nosignificantincreaseswereseeninanytumorfromthisstudy.However,giventhesmallsamplesizeforhistopathologicalevaluationandthelowdosesusedforthisstudy,thisstudyisinadequate.

Thisstudywillnotbeusedintheevaluationofcausality.

KnezevichandHogan,(1983)[83]exposedCD-1micetoglyphosate(99.8%pure)infeedfortwoyears.Fiftyanimals/group/sexweretestedinfourexposuregroups(seeTable9).

TherewerenosurvivaldifferencesinthisstudyandtherewasnoindicationthatthehighestdoseusedexceededtheMTD.

EPA[100]foundasignificantincreaseinkidneytubularcelladenomasinmalemicebasedupontheoriginalpathologydonefromthestudyandthisanalysisisshowninTable9(pTrend=0.019).KidneytubularcelladenomasareveryraretumorsinCD-1micesoitisimportanttocomparetheseresultswiththehistoricalcontrols.NohistoricalcontrolswereavailablefromthelaboratorythatconductedKnezevichandHogan,(1983)soIARC,EPAandEFSAallusedhistoricalcontroldatabasesfrompublishedstudiesinthe


37

literature[101-103].ThesestudieshavevirtuallyidenticalratesfortheimportanttumorsseeninCD-1mice;IwillusethestudybyGiknisandClifford(2000)[102]sinceitbestcoverstherangeofstudieswehaveforCD-1mice.Forstudiesofapproximatelytwoyears,themeanhistoricaltumorresponseincontrolsis0.27%.ApplyingthiscontrolresponseratetothekidneyadenomasyieldspHist=0.005,strengtheningthesignificanceoftheevaluationagainsttheconcurrentcontrol.EPAoriginallyusedasimilaranalysisandreachedthesameconclusions.However,in1985,theregistranthadagroupofpathologistsreviewthekidneyslides.Usingadditionalkidneysectionsfromthisstudy,thepathologistsidentifiedanadditionaladenomainthecontrolanimalsandchangedtheclassificationforthreeadenomastocarcinomas(Table9).Withthesechanges,theadenomasnolongerhaveasignificanttrend(PTrend=0.442,PHist=0.121)butcarcinomashaveamarginallysignificanttrendagainstconcurrentcontrolsandaclearlysignificanttrendusinghistoricalcontrols(pTrend=0.063,pHist=0.002,historicalcontrolrateof0.15%).Thesehistoricalcontrolratesmaynotapplytothisanalysisbecausethereevaluationofthekidneytumorsconsideredadditionalsectionsandnoinformationisavailableonhowadditionalsectionsaffecthistoricalcontrolratesinthisstrainofmice;differenceshavebeenseeninothersettings[104].Theincidenceofcombinedcarcinomasandadenomashasthesamemarginalsignificanceagainsttheconcurrentcontrolandsignificanceagainstthehistoricalcontrols(pTrend=0.065,pHist=0.011,historicalcontrolrateof0.44%).However,therewasconsiderabledisagreementonwhethertheoneadenomainthecontrolgroupwascorrectlydiagnosed[105].RemovingthisoneadenomafromthecontrolgroupresultsinpTrend=0.019andpHist=0.005.

OtherCD-1mousestudieshaveseenincreasesinmalignantlymphomas,hemangiosarcomasandlungadenocarcinomas(males)andhemangiomas(females).Evaluationsofthosetumorsforthisstudyyieldsresultsthatarenotsignificant;formalignantlymphoma,pTrend=0.754,pHist=0.767,withthehistoricalcontrolrateequal6.2%,forhemangiosarcomaspTrend=0.503,pHist=0.591,withthehistoricalcontrolrateequalto2.5%,forlungadenocarcinomaspTrend=0.918,pHist=0.899,withthehistoricalcontrolrateequalto9.2%andforhemangiomaspTrend=0.631.Noothertumorswerefoundinthisstudy.

TheEPA[61]hasproducedmanydifferentargumentstodismissthefindingsofrenaltumorsfromthisstudy.OneargumentisthatthepathologyworkinggrouprequestedbytheEPAin1986concludedtheselesionswerenotglyphosaterelatedbecause“1)renaltubularcelltumorsarespontaneouslesionsforwhichthereisapaucityofhistoricalcontroldataforthismousestock;2)therewasnostatisticalsignificanceinapairwisecomparisonoftreatedgroupswiththeconcurrentcontrolsandtherewasnoevidenceofastatisticallysignificantlineartrend;3)multiplerenaltumorswerenotfoundinanyanimal;and4)compound-relatednephrotoxiclesions,includingpre-neoplasticchanges,werenotpresentinmalemiceinthisstudy.”ReasonnumberonenolongerexistsastherearetwoverygoodhistoricalcontroldatabasesforCD-1mice[101,102].Thesecondreason,whiletechnicallycorrect,isnotsupportablesincetheAgency’sownguidelinesforevaluatingcarcinogenicitystudiesstatethat“Significancein


38

eitherkindoftest[trendorpair-wise]issufficienttorejectthehypothesisthatchanceaccountsfortheresult.”Thethirdreasonisalsoweaksinceonewouldnotexpect(norrequire)multipletumorstoappearwhendealingwithararetumor.Forthefourthpoint,EPAprovidesdataontherateofbilateralchronicinterstitialnephritisinthestudywhichitconsiderstoshownostatisticallysignificantresultsalthoughthetrendtestishighlysignificant(pTrend=0.006,Table9).EPAthenstates,withoutreference,that“chronicinterstitialnephritisisnotconsideredtobeaprecursorlesionfortubularneoplasms”.Icouldfindnopublishedresearchtoeithersupportorrefutethisstatement.However,chronicinterstitialnephritisisaninflammationoftheinterstitialtissuesurroundingtheglomeruliandtubulesinthekidney.Inflammationiswellknown

toplayanimportantroleinkidneycancer[106]andmanyothercancerssothisargumentalsofailstosupportrejectionofthesefindings.

Table9:TumorsofinterestinmaleandfemaleCD-1micefromthe24-monthfeedingstudyofKnezevichandHogan(1983)[83]


Male 0 157 814 4841Female 0 190 955 5874

KidneyAdenoma1(originalpathology)

Male 0/49 0/49 1/50 3/50 PTrend=0.019PHist=0.005

KidneyAdenoma(EPApathology)


KidneyCarcinoma2(EPApathology)u


KidneyAdenomaandCarcinomaCombined3(EPApathology)


MalignantLymphoma4 Male 2/49 5/49 4/50 2/50 PTrend=0.754PHist=0.767

Hemangiosarcoma5 Male 0/50 0/49 1/50 0/50 PTrend=0.503PHist=0.591

BilateralChronicInterstitialNephritis

Male 5/49 1/49 7/50 11/50 PTrend=0.006

Hemangiooma6 Female 0/49 1/49 1/50 0/50 PTrend=0.631

LungAdenocarcinoma7 Male 4/48 3/50 2/50 1/50 PTrend=0.918PHist=0.899

*-pFisher<0.05,**-pFisher<0.01,1historicalrate=0.27%,2historicalrate=0.15%,3historicalrate=0.44%,4historicalrate=6.2%,5historicalrate=2.5%,6NoHistoricalControls,7Historicalrate=9.2%


39

Insummary,thisstudyshowsapositiveresultforkidneytumorsinmaleCD-1miceandwillbeincludedintheoverallevaluationofcausation.

Atkinson,etal.,(1993)[81]exposedCD-1micetoglyphosate(>97%purity)infeedfortwoyears.Fiftyanimals/group/sexweretestedinfourexposuregroups(seeTable10).

TherewasnoimpactonsurvivalofadministrationofglyphosateandnoindicationthatthehighdoseexceededtheMTD.

Table10:TumorsofinterestinmaleandfemaleCD-1micefromthe24-monthfeedingstudyofAtkinsonetal.(1993)[81]


Male 0 98 297 988Female 0 102 298 1000

KidneyAdenomaandCarcinomaCombined1

Male 2/50 2/50 0/50 0/50 PTrend=0.981PHist=1



Hemangioma4 Female 0/50 0/50 0/50 0/50 PTrend=1


*-pFisher<0.05,**-pFisher<0.01,1historicalrate=0.44%,2historicalrate=6.2%,3historicalrate=2.5%,4Nohistoricalcontrolrate,5Historicalrate=9.2%

Hemangiosarcomasweretheonlytumorsshowingasignificanttrendinthisstudy(PTrend=0.004,PHist=0.001,Table10).AlsoshowninTable10aretheresultsformalignantlymphomas,kidneytumorsandlungadenocarcinomas(males)andhemangioma(females);thereisamarginaltrendformalignantlymphomas(PTrend=0.087,PHist=0.085)andnotrendforkidneytumors.

TheEPA[61]concludedthefindingsinthisstudywerenottreatmentrelatedbaseduponthetumorsappearingonlyinthehighdosegroup,alackofstatisticalsignificancebetweentheresponseinthisgroupandcontrolresponseandthatthesetumorsarecommonlyobservedinmiceasbothspontaneousandtreatmentrelatedeffects.Thereisnoscientificsupportforexcludingpositivefindingsinthehighestdosegroup,aviewalsoheldbytheSAP[54].IhavealreadycommentedonhowEPA’sguidelinestreattrendtestsandFisher’sExacttestresults,althoughinthiscase,thevalueofthecomparisonofthehighestexposuregrouptocontrols,pFisher=0.059,ismarginallysignificant.Theargumentregardingthefrequencyofthistumorincontrolsisaddresseddirectlybytheevaluationagainstthehistoricalcontrolrates;iftheserateswerehighenoughtoexcludethisfinding,PHistwouldhavebeabove0.05insteadof0.001.Themean


40

historicalcontrolincidenceofhemangiosarcomasincontrolsfromtwo-yearcancerbioassaysinCD-1miceis2.5%andtheresponseseeninthehigh-dosegroupis8.9%.TheSAP[54]statedveryclearlythatthepractice,beingusedbytheEPA,ofnegatingapositivefindingbecauseofhistoricalcontroldatawasnotacceptable[54].(page63).TheEPACancerGuidelines[33]statethisveryclearly“…statisticallysignificantincreasesintumorsshouldnotbediscountedsimplybecauseincidenceratesinthetreatedgroupsarewithintherangeofhistoricalcontrolsorbecauseincidenceratesintheconcurrentcontrolsaresomewhatlowerthanaverage.”

Insummary,thisstudyshowsapositiveresultforhemangiosarcomasinmaleCD-1miceandwillbeincludedintheoverallevaluationofcausation.

Woodetal.,(2009)[88]exposedCD-1micetoglyphosate(95.7%pure)infeedfor80weeks.Fifty-oneanimals/groups/sexweretestedinfourexposuregroups(seeTable11).

TherewasnoeffectonsurvivalandnoinformationsuggestingthestudyexceededtheMTD.

Noincreaseinkidneytumorsorhemangiosarcomas(males)orhemangiomas(females)wereseeninthisstudy.Therewasamonotonicincreaseinlungadenocarcinomas(pTrend=0.028,pHist=0.031)inmalesandamonotonicincreaseinmalignantlymphomas(pTrend=0.007,pHist=0.007)inmales.Thehistoricalcontrolincidenceforthisstudyisdifferentfromtheearlierstudiesbecausethisstudyisonlyfor80weeksinsteadof104weeks(twoyears);thehistoricalcontrolrateformalignantlymphomasinCD-1miceafter80weeksis2.6%insteadof6.2%,thehistoricalcontrolrateattwoyears[102].

Forlungadenocarcinomas,theEPA[61]againarguedalackofsignificanceforpairwisecomparisons(inviolationofitsguidelines)andthattherewasnoevidenceofprogressionfromadenomastocarcinomas.Eventhoughtherewasnoincreaseinlungadenomasasafunctionofexposure,itispossibletohaveanincreaseinlungadenocarcinomaswithoutanassociatedincreaseinadenomas[107].Formalignantlymphomas,EPAnotesthattherewasastatisticallysignificantresponseandthatthehighdosewassignificantlydifferentfromcontrol(pFisher=0.028),butthenusesanargumentbaseduponthenumberofanalysesdoneinthisstudytoadjusttheFisherExacttestp-valueto0.082(anadjustmentformultiplecomparisonsisindeedwarrantedinevaluatingtheoutcomesoftheseanimalcancerstudies,thiswillbeaddressedlaterinmyreportintheevaluationofallofthestudiescombined).

TheEPA[61]useshistoricalcontroldata[103,108]toexcludethemalignantlymphomasandciteameanresponseof4.5%andarangeof1.5%to21.7%.SonandGopinath(2004)[108]saw21animalsoutof1453examinedpriorto80weekswithlungadenocarcinomas(1.4%).GiknisandClifford(2005)[103]sawameanrateof4.5%witharangeof0%to21.7%in52studieswhichincludedmostly78weekcontrols(26studies)and104weekcontrols(21studies).Includingonlystudiesof80weeksorless,therateinGiknisandClifford(2005)is37/1372=2.7%witharangeof0%to14%.GiknisandClifford(2000)[102](thereferenceIhavebeenciting)didasimilarevaluation,usingmostlythesamedataastheir2005paperandsawanaveragetumorincidencebefore


41

80weeksof2.6%witharangeof0%to14%.BaseduponitsflawedinterpretationoftheGiknisandClifford(2005)historicalcontrols,EPAarguesthattheincidenceofconcurrentcontrolsinthestudywaslow(itwas0%)andrejectedthepositivefinding.Infact,ofthe26studiesinthe18-monthcontrolgroupsevaluatedbyGiknisandClifford(2005),eight(31%)hadresponseof0%andeight(31%)hadonlyonetumor.TheevaluationusedbytheEPAisincorrect.Inaddition,asnotedearlier,theuseofhistoricalcontroldatatonegateapositivefindingisnotsupportedbyEPA’sguidelines[33,54]oritsSAP[54].

Therewasanincreaseinthenumberofanimalswithmultiplemalignanttumors(PTrend=0.046)

Insummary,thisstudyshowsapositiveresultformalignantlymphomasandlungadenocarcinomasinmaleCD-1miceandwillbeincludedintheoverallevaluationofcausation.

Table11:TumorsofinterestinmaleandfemaleCD-1micefromthe18-monthfeedingstudyofWoodetal.(2009)[88]


Male 0 71.4 234.2 810Female 0 97.9 299.5 1081.2

KidneyAdenoma1 Male 0/51 0/51 0/51 0/51 PTrend=1MalignantLymphoma2 Male 0/51 1/51 2/51 5/51* PTrend=0.007

PHist=0.007Hemangiosarcoma Male 0/51 0/51 0/51 0/51 PTrend=1

LungAdenocarcinoma3 Male 5/51 5/51 7/51 11/51 pTrend=0.028PHist=0.031

Hemangioma4 Female 0/51 2/51 0/51 1/51 pTrend=0.438

AnimalswithMalignantNeoplasms

Male 14/51 20/51 17/51 20/51 PTrend=0.203

AnimalswithMalignantNeoplasms

Female 23/51 15/51 17/51 18/51 PTrend=0.628

Animalswithmultiplemalignanttumors

Male 1/51 2/51 3/51 5/51 PTrend=0.046

*-pFisher<0.05,**-pFisher<0.01,1historicalrate=0.44%,2historicalrate=2.6%,3Historicalrate=2.5%,4NoHistoricalControlRate

Sugimoto(1997)[87]exposedCD-1micetoglyphosate(94.61-95.67%pure)infeedfortwoyears.Fiftyanimals/group/sexweretestedinfourexposuregroups(seeTable12).

TherewerenoeffectsoftreatmentonsurvivalandnoindicationthehighestdosehadexceededtheMTD.

Kidneyadenomas(pTrend=0.062,pHist=0.005),malignantlymphomas(pTrend=0.016,


42

pHist=0.017)andhemangiosarcomas(pTrend=0.062,pHist=0.004)inmalemiceandhemangiomas(pTrend=0.002)infemalemiceallshowedincreasedtumorincidencewithincreasingdose.Theevaluationoflungadenocarcinomasinmalesshowednosignificantdose-relatedtrend(pTrend=0.148,pHist=0.140).Thisstudyalsohadanincreaseinanimalswithanymalignancyinmales(pTrend=0.001)butnotinfemales(pTrend=0.362).Notethatnohemangiosarcomaswereseeninthe26controlgroupsevaluatedbyGiknisandClifford(2000)sothedevelopmentofanestimateofthehistoricalcontrolresponseisdifficult(ifthehistoricalcontrolrateis0,thenanyobservedresponseotherthan0hasap-valueof0).Thefactthatthistumorwasneverseeninthehistoricalcontrolsshouldstronglysupportanypositivefindingasbeingsignificant.However,tostillallowforatestusinghistoricalcontroldata,Iusedthehistoricalcontrolestimateofthemeanresponsethatwouldresultina5%chanceofseeingnotumorsin1149animals.Thisestimatedhistoricalcontrolresponsevaluewas0.0026.ThisvaluewasusedintheanalysisforhemangiosarcomasinmaleCD-1miceexposedfor18months(pHist<0.001).

Table12:TumorsofinterestinmaleandfemaleCD-1micefromthe18-monthfeedingstudyofSugimoto(1997)[87]


Male 0 165 838.1 4348Female 0 153.2 786.8 4116

KidneyAdenoma1 Male 0/50 0/50 0/50 2/50 PTrend=0.062PHist=0.005



Hemangioma4 Female 0/50 0/50 2/50 5/50* PTrend=0.002


NumberofanimalswithMalignantNeoplasms

Male 5/50 5/50 11/50 16/50** PTrend=0.001

NumberofanimalswithMalignantNeoplasms

Female 9/50 13/50 16/50 13/50 PTrend=0.362

*-pFisher<0.05,**-pFisher<0.01,1historicalrate=0.44%,2historicalrate=2.6%,3historicalrate=0/1424(0.26%-95%confidencelimit),4NoHistoricalControlRate,5Historicalrate=2.5%

EPA[61]onlyaddressedthehemangiomasinthefemalemiceanddidnotnoteanyothersignificanteffects.Forthefemales,EPAarguedthatthehighdosewasapproximatelyfourtimeshigherthanthecurrentrecommendedhighdosefromtheOECDguidelines[109].Thisstudywascorrectlydesignedunderthepreviousguidelines(thelimitwas<5%infeed)andthereisnoindicationthatthisdoseexceededtheMTD.TheEPAalsoarguedthatwhenthep-valueforFisher’sExacttestwasadjustedformultiplecomparisons,thenewp-valueforthehigh-dosegroupforhemangiomaswas0.055.


43

Forthehemangiosarcomasinmales,noneofthe26historicalcontrolgroupsexaminedbyGiknisandClifford(2000)hadhemangiosarcomas,makingthisaveryraretumorinmalespriorto80weeksonstudy.Themalignantlymphomasinmalesarestatisticallysignificantagainstboththeconcurrentcontrolsandthehistoricalcontrols.Finally,thereisclearlyanoverallincreaseofmalignanciesinthemales.

Insummary,thisstudyshowsapositiveresultforkidneyadenomas,malignantlymphomasandhemangiosarcomasinmaleCD-1mice,hemangiomasinfemaleCD-1miceandanoverallincreaseinmalignanciesasafunctionofexposureinmaleCD-1mice.Thisstudywillbeincludedintheoverallevaluationofcausation.

Kumar(2001)[84]exposedSwissAlbinomicetoglyphosate(>95%purity)infeedfortwoyears.Fiftyanimals/group/sexweretestedinfourexposuregroups(seeTable13).

ThesurvivalwasdecreasedinthehighestexposuregroupbutthiswasnotstatisticallysignificantandtherewasnootherdataindicatingtheMTDwasexceededforthisstudy.

Kidneyadenomas(pTrend=0.062)andmalignantlymphomas(pTrend=0.064,pHist=0.070)inmalemicedemonstratedmarginalstatisticalsignificanceandhemangiosarcomas(pTrend=0.500)inmalemicedemonstratednostatisticalsignificance.Inthisstudy,notallanimalsinthelow-andmid-dosegroupswereevaluatedforkidneytumors,soasecondanalysiswasdonebasedononlytheanimalsexaminedinthesetwogroups(pTrend=0.088).NohistoricalcontroldatawasavailableforhemangiosarcomasandkidneyadenomasinSwissAlbinomice.Forthemalignantlymphomas,EFSAprovidedahistoricalcontroldatasetshowingameanresponseof46/250=0.184(18.4%)witharangeof6%to30%.Usingthishistoricalcontroldata,thetrendisonlymarginallysignificant(pHist=0.070).Ihavesomeconcernthattheresponsesattwoofthedosesareoutsideofthehistoricalcontrolrangeandthethirddoseisattheupperlimitofthehistoricalcontrolrange.However,thisisasmallhistoricalcontroldatasetforatumorwitharelativelyhighbackgroundtumorrate,thusplacingtoomuchemphasisonthishistoricalcontrolpopulationisnotwarranted.

Inarecentmemo,Martens(2017)[110]assertsthattheincidencecountsformalignantlymphomasandkidneyadenomasappearinginGreimetal.(2015)[91]andEFSA(2013)[89]areincorrectandprovidesdifferentrates(showninTable13).Thep-valuesforbothofthesetumorsarereducedusingtheincidencecountsfromtheMartensmemo.However,itshouldbenotedthatifthecountsformalignantlymphomasintheMartens(2017)memoarecorrect,thenallthreeexposuregroupshaveresponsesoutsideoftherangeofthehistoricalcontrols.ItisunclearfromGreimetal.(2015),EFSAorMartens(2017)whichtumorincidencecountsarecorrect.

Therewasasignificantincreaseinhemangiomas(anytissue)infemalemice(pTrend=0.004).

Insummary,thisstudyshowssupportforanincreaseformalignantlymphomasandkidneyadenomasasafunctionofexposureinmaleSwissAlbinomiceandanincreaseinhemangiomasinfemaleSwissAlbinomice.Thisstudywillbeincludedintheoverall


44

evaluationofcausation.

PavkovandTurner(1987)[85]exposedCD-1micetoglyphosatetrimesiumsalt(56.2%)and1%propyleneglycol(wetweightvehicle)infeedfortwoyears.Eightyanimals/sex/groupweretestedincontrol,low-andmid-dosegroupsand90animals/sexweretestedatthehighdose.Exposurelevelswere0,11.7,118and991mg/kg/dayinmalesand0,16,159and1341mg/kg/dayinfemales.EPA[61]liststhisstudyascompletelynegativeforanycancerfindings.NodetailsonthisstudyareprovidedbytheEPAnorisitlistedintheGreimetal.(2015)[91]manuscript.TherewaslimitedinformationonthisstudyinaDataEvaluationReportfromEPA(accessionnumber402140-06)thatdiscussedfindingsfromthisstudy.EPAnotedthatbodyweightandfoodconsumptionwerereducedinthehighestexposuregroup,buttheactualamountsofthesereductionswerenotavailable.Theyalsonotedthattheauthorsfailedtomakeitclearthatthetumorsreportedinthestudyhadbeenhistopathologicallyvalidated.Datawaspresentedfortumorsintheliversandlungsofmalemiceandthelungsoffemalemice.Nootherdataisprovided.

Thisstudyisnotacceptableforinclusionintheevaluationofcausationduetothelackofinformationonthetumorincidenceintissuesotherthanliverandlung.

Georgeetal.(2010)[82]exposedgroupsof20maleSwissAlbinomicetoaglyphosate

Table13:TumorsofinterestinmaleandfemaleSwissAlbinomicefromthe18-monthfeedingstudyofKumar(2001)[84]


Male 0 14.5 149.7 1453Female 0 15 151.2 1466.8

KidneyAdenoma(onlytissuesexaminedmicroscopically)

Male 0/50 0/26 1/22 2/50 PTrend=0.088

KidneyAdenoma(asreportedbyGreimetal.)

Male 0/50 0/50 1/50 2/50 PTrend=0.062

KidneyAdenoma(asreportedbyMartens)

Male 0/50 0/50 0/50 1/50 PTrend=0.250

MalignantLymphoma1(asreportedbyGreimetal.)


MalignantLymphoma1(asreportedbyMartens)

Male 10/50 16/50 18/50 19/50* PTrend=0.141PHist=0.150

Hemangiosarcoma Male 0/50 0/50 2/50 0/50 PTrend=0.500

Hemangioma(anytissue) Female 1/50 0/50 0/50 5/50 PTrend=0.004

*-pFisher<0.05,**-pFisher<0.01,1Historicalcontrolrate=0.184(46/250mice)


45

formulation(RoundupOriginal,36g/Lglyphosate)atadoseof25mg/kg(glyphosateequivalentdose)topicallythreetimesperweek,topicallyoncefollowedoneweeklaterby12-o-tetradecanoylphorbol-13-acetate(TPA)threetimesperweek,topicallythreetimesperweekforthreeweeksfollowedoneweeklaterbyTPAthreetimesperweek,orasingletopicalapplicationof7,12-dimethyl-benz[a]anthracene(DMBA)followedoneweeklaterbytopicalapplicationofglyphosatethreetimesperweekforatotalperiodof32weeks.Appropriateuntreated,DMBA-treated,andTPA-treatedcontrolswereincluded.ThegroupexposedtoDMBAfollowedbyglyphosatedemonstratedasignificantincrease(p<0.05)inthenumberofanimalswithtumors(40%ofthetreatedanimalsversusnotumorsinthecontrols)indicatingglyphosatehasapromotionaleffectoncarcinogenesisinthetwo-stagemodelinskin.Thisstudyaddressesthequestionofwhetherglyphosateismorelikelytocauseskintumorsthroughinitiation(startingthecancerprocess)orpromotion(movingtheprocessalongafteritstarts).Thisstudysupportstheoverallconceptthatglyphosatecanhaveanimpactontumorincidence.

EPA[61]discountedthisstudybecauseitincludedonly20animalspergroup,testedonlymalesanddidnotconductahistopathologicalanalysis.ItishardtounderstandhowEPAcouldrejectapositivefindingusing20mice;typicallyonewouldignoreanegativestudythathadtoofewanimalsasnothavingsufficientstatisticalpowertoseeaneffectbutneverrejectpositivefindingsforthisreason.Also,20animalspergroupiscommonforskin-paintinginitiation-promotionstudiesliketheonepresentedhere.Doingastudyinonlymalesisnotareasontoignorethepositivefindingsinastudy.Finally,ininitiation-promotionstudiesofmouseskin,histopathologicalevaluationwouldbedoneifonewereinterestedinseparatingpapillomasfromcarcinomas.Itishighlyunlikelythatthelesionsseenin40%oftheDMBA/glyphosatetreatedmicewerenotpapillomasorcarcinomas.

SomemembersoftheEPASAPnoted[54]thattherodentdatawereconsistentwithglyphosateactingasatumorpromoterbut,because“[t]herehasbeennodirecttestofthishypothesis(suchasinastandardinitiation-promotionbioassay)…,”this“conclusionwasspeculative.”(page#).BecausetheEPAdismissedthisstudywithoutanydiscussion,theSAPdidnotrecognizetherewasaninitiation-promotionsupportingapromotionaleffectofglyphosate.

Thisstudyisincludedintheevaluationofcausalityassupportforapromotionaleffectofglyphosateonsometumors.

JointAnalysis-Mouse

Intheirevaluationofthemousestudies,EPA[61]andEFSA[89]chosetochallengetheresultsineachstudyseparately,dismissthestudiesasshowingnoeffect,andnevercomparedresultsacrossthevariousstudies.InresponsetotheevaluationdonebytheIARC[30],EFSA[90]extractedtheoriginaldataanddidtrendtestsonkidneytumors,malignantlymphomasandhemangiosarcomasinmalemiceinfiveofthemousestudies,thesamefivestudiesIconsideracceptableforacausationanalysis.Ratherthanformallyevaluatethesecancerresponsesforconsistencybypoolingthedatawhereappropriate,EPAandEFSAsimplyproducedatablewiththeresponsesforeachdose


46

groupineachstudyandconcluded(subjectively)theywereinconsistent.Inaddition,EPAandEFSAarguedthatdosesabove1000mg/kg/day(thereareonlytwoofthese)wereoutsidetherangeofwhatwouldbetestedtodayunderOECDguidelinesandshouldbeexcluded.Iwillnowaddressbothpoints.

InCD-1mice,therearefourusefulanimalcarcinogenicitystudiesandonestudyinSwissAlbinomice.Aswiththerats,consistencyacrossstudiescanbeaddressedintwoways.Thefirstisbysimplylookingattheoverallfindingstoevaluatewheretheyagreeordisagreeintermsofstatisticalsignificance.Table14summarizesthepositiveandnegativefindingsforallfivecancersinwhichatleastonestudyinCD-1miceshowedasignificanttrend.Itisclearthatnoteverytumorshowsapositivetrendwithglyphosateexposureineverystudy.Forhemangiosarcomasinmales,thereareclearpositivefindingsinthestudiesbySugimoto(1997)andAtkinsonetal.(1993)andnon-significantresponsesinWoodetal.(2009)andKnezevichandHogan(1983).Infemales,hemangiosarcomasareonlypresentinthestudybySugimoto(1997).Malignantlymphomasinmalesareclearlypositiveintwostudies[87,88]andmarginallypositiveinathird[81]butnegativeinthefourth[83].Bothofthestrongpositivestudiesexposedanimalsfor18months.Kidneytumorsinmalesarepositiveintwostudies[83,87]andnegativeintheremainingtwo[81,88].LungadenocarcinomasinmalesareonlypositiveinthestudybyWoodetal.(2009).Sugimoto(1997)hadfourclearlypositiveassociationsbetweentumorsandglyphosatewhiletheothershadtwoorless.

Table14:Summaryofsignificancetestsfor5tumorsfrom4studiesinCD-1Mice

Study

Monthson

Study

Neoplasm

Hemangio-sarcoma(male)

Hemangioma(female)

MalignantLymphoma(male)

KidneyTumor(male)

LungAdeno-carcinoma(male)

Sugimoto1997[87]

18 +/+++1 +++ ++/++ +/+++ -/-

Wood2009[88]

18 -/- - +++/+++ -/- ++/++

Sugimoto&WoodPooled

++/+++ +++ +++/+++ ++/+++ -/-

Atkinson1993[81]

24 +++/+++ - +/+ -/- -/-

Knezevich1983[83]

24 -/- - -/- +/++ -/-

Atkinson&KnezevichPooled

-/- - -/- +/+ -/-

AllCD-1StudiesPooled

++/++ ++/++ +/+ +++/+++ -/-

1entriesarepTrend/pHistwithvalues:–p>0.1,+0.1≥p>0.05,++0.05≥p>0.01,+++p≤0.01


47

Asseenfortheratstudies,thissimpleevaluationofthepositiveversusnegativefindingsfailstoresolvetheissueofwhichfindingsaredrivingtheoverallresponsesinthesedata.Todothis,Iwillagainpoolthestudies.Table14summarizesthepooledanalyses.

Forkidneytumorsinmales,poolingthetwo18-monthstudiesyieldssignificantincreasesinincidence(pTrend=0.015,pHist=0.003)andpoolingofthetwoyearstudiesshowsmarginalsignificance(pTrend=0.081,pHist=0.054).Poolingallfourstudiesresultsin(pTrend=0.005,pHist=0.007),thusthepositivetrendremains.KnezevichandHogan(1983)sawa4%responseforkidneycarcinomasintheirhighestexposuregroup.Thelargestresponseseenforkidneycarcinomasincontrolsin48studiesbyGiknisandClifford(2000)andin52studiesbyGiknisandClifford(2005)was2%andinthecontrolgroupsfrom11two-yearcancerstudies,ChandraandFrith(1992)[101]sawonlyoneanimaloutof725withakidneycarcinoma.In46controldatasets,GiknisandClifford(2000)saw39controlgroupswithnoadenomas,fivewithoneadenomaandtwowithtwoadenomas;both24-monthstudiessawtwoadenomasinthehighestexposuregroup,averyrarefinding.Tobetterillustrate,thereare16groupsofanimalsinthefourstudies.Foranyonegroup,thereisa2/44or4.3%chanceofgettingaresponse4%orlarger.Thechancesofrandomlygetting3ormoresuchresponsesin16groupsis2.9%andthechancesoftwoofthesebeinginanytwoofthefourhighestexposuregroupsis0.01.Insummary,thestrongfindingintwoofthefourstudies,thepositivefindingwhenallfourstudiesarepooledandtheverylowprobabilitythatthisisduetochancewhencomparedtohistoricalcontrolssupporttheconclusionthatglyphosatecauseskidneytumorsinmalemice.

Formalignantlymphomasinmales,poolingthetwo18-monthstudies,Sugimoto(1997)andWoodetal.(2009),resultsinasignificanttrend(pTrend=0.005,pHist=0.006).Poolingthetwo24-monthstudies,KnezevichandHogan(1983)andAtkinsonetal.(1993),yields(pTrend=0.653,pHist=0.649).Themaindifferencesbetweenthesetwofindingsisinthecontrolresponse;thepooledcontrolresponseat24monthsis6/99(6%)versus2/101at18months(2%).Thisisexpectedsince,intheabsenceofanyexposure,tumorratesincreaseasafunctionofage[5].GiknisandClifford(2000)showacontrolresponseat18monthsof4%andacontrolresponseat24monthsof6%(matchingthevalueforthepooledstudies).Poolingallfourstudiesresultsin(pTrendA=0.073,pHist=0.080).However,theresponsesseenformalignantlymphomasincontrolsbyGiknisandClifford(2000)showonlyonehistoricalcontrolgroupintwenty-six18-monthgroupswith10%orhigherresponse.Theresponsesatthehighdoses(10%and12%)inthetwo18-monthstudiesareveryunlikelytohavearisenbychance.Thereareeightgroupsofanimalsinthetwostudies.Foranyonegroup,thereisa1/26or3.8%chanceofgettingaresponseofatleast10%basedonthe26controlgroupsfromGiknisandClifford(2000).Thechancesofgettingtwoormoresuchresponsesineightgroupsis0.035andthechancesofthesebeinginthreeofthefourhighestexposuregroupsis0.004.Forthe24-monthstudies,thehigherbackgroundratemakesitdifficulttoidentifyasmallchangeinincidence,thusthefindingsinthe24-monthstudiesandthe18-monthstudiesarenotinconsistent.Insummary,theverystrongfindingsinthe18-monthstudies,theverystrongpositivefindingswhenthetwo18-monthstudiesarepooled,thelowprobabilitythattheresponsesseeninthe18-monthstudiesareduetochance,andthe


48

marginalincreaseinmalignantlymphomasinthe18-monthstudyinSwissAlbinomice[84]supporttheconclusionthatglyphosatecausesmalignantlymphomainmalemice.

Forhemangiosarcomasinmales,poolingthetwo18-monthstudiesresultsinasignificanttrend(pTrend=0.015,pHist=0.002).Poolingthetwo24-monthstudiesyields(pTrend=0.490,pHist=0.429).Themaindifferencebetweenthesetwofindingsisthe0/50responseinanimalsexposedat4841mg/kg/dayinthestudybyKnezevichandHogan(1983).Removingthisoneexposuregroupinthepooled24-monthanalysisyields(pTrend<0.001,pHist<0.001).Poolingallfourstudiesresultsin(pTrend=0.045,pHist=0.043).Nohemangiomaswereseenincontrolsgroupsfromtwenty-six18-monthstudiesbyGiknisandClifford(2000)sothetwohemangiosarcomasseeninthehighdosegroupinthestudybySugimoto(1997)arebiologicallyverysignificant.Forthe24-monthhistoricalcontrols,onlytwooutof20controlgroupshadaresponsegreaterthan8%.Insummary,theverystrongfindingsinthe18-monthstudies,thepositivefindingwhenallfourstudiesarepooledandthelowprobabilitythattheresponsesseeninthe18-monthstudiesareduetochancesupporttheconclusionthatglyphosatecauseshemangiosarcomasinmaleCD-1mice.

Forhemangiomasinfemales,poolingthetwo18-monthstudiesresultsinasignificanttrend(pTrend=0.001).Poolingthetwo-yearstudiesresultsinpTrend=0.424.PoolingallfourstudiesresultsinpTrend=0.018.Insummary,theverystrongfindingsinone18-monthstudy,thepositivefindingwhenallfourstudiesarepooledandthelowprobabilitythattheresponsesseenintheSugimoto(1997)studyareduetochance,supporttheconclusionthatglyphosatecauseshemangiomasinfemaleCD-1mice.

ForlungadenocarcinomasinmaleCD-1mice,poolingthetwo18-monthstudiesresultsshowsnosignificanttrend(pTrend=0.417,pHist0.126).Poolingthetwo24monthstudiesyields(pTrendA=0.985,pHist=0.993).Poolingallfourstudiesresultsin(pTrendA=0.937,pHist=0.744).Insummary,themoderatefindingsinone24monthstudy,andthenegativefindingwhenanystudiesarepooledsuggestthatthelinkagebetweenglyphosateandlungadenocarcinomasinmaleCD-1miceisduetochance.

TheonestudyinSwissAlbinomice[84]waseffectivelynegativeforallendpointsexceptmalignantlymphomasandkidneyadenomaswheremarginallysignificanttumorresponseswereseen.ConsideringthefindingsforkidneyadenomasinCD-1mice,glyphosatemayalsocausekidneyadenomasinmaleSwissAlbinomicefromthestudyofKumar(2001).

Tosummarizethefindingsinmice,glyphosatecauseshemangiosarcomas,kidneytumorsandmalignantlymphomasinmaleCD-1miceandhemangiomasinfemaleCD-1miceafter18monthsofexposure,kidneytumorsinmaleCD-1miceafter24monthsexposureandpossiblykidneyadenomasinmaleSwissalbinomice.When18-monthand24-monthstudiesarepooled,thereisasignificantincreaseinhemangiosarcomasinmalemice,hemangiomasinfemalemiceandkidneytumorsinmalemice.

DiscussionandSummaryAnimalCarcinogenicityStudies


49

Asnotedearlier,therehasbeenasuggestionthatusingdosessubstantiallylargerthan1000mg/kg/dayexceedsthecurrentlimitdosesetbytheOECD.TheonlyplaceintheOECDguidance[67]thataddressesadoseof1000mg/kg/dayisinparagraph23whichreads:

“Forthechronictoxicityphaseofthestudy,afullstudyusingthreedoselevelsmaynotbeconsiderednecessary,ifitcanbeanticipatedthatatestatonedoselevel,equivalenttoatleast1000mg/kgbodyweight/day,isunlikelytoproduceadverseeffects.Thisshouldbebasedoninformationfrompreliminarystudiesandaconsiderationthattoxicitywouldnotbeexpected,basedupondatafromstructurallyrelatedsubstances.Alimitof1000mg/kgbodyweight/daymayapplyexceptwhenhumanexposureindicatestheneedforahigherdoseleveltobeused.”

Thislanguagedoesnotprecludetheuseofadoseexceeding1000mg/kg/daynordoesitadvocateignoringsuchdoseswhenevaluatingtheresultsofananimalcarcinogenicitystudy.Infact,thereasonsforexcludingadoseinananimalcarcinogenicitystudyareclearlyoutlinedinparagraph90withinOECDguidance[59]andreads:

“Ifthemainobjectiveofthestudyistoidentifyacancerhazard,thereisbroadacceptancethatthetopdoseshouldideallyprovidesomesignsoftoxicitysuchasslightdepressionofbodyweightgain(notmorethan10%),withoutcausinge.g.,tissuenecrosisormetabolicsaturationandwithoutsubstantiallyalteringnormallifespanduetoeffectsotherthantumours.Excessivetoxicityatthetopdoselevel(oranyotherdoselevel)maycompromisetheusefulnessofthestudyand/orqualityofdatagenerated.Criteriathathaveevolvedfortheselectionofanadequatetopdoselevelinclude:(inparticular)toxicokinetics;saturationofabsorption;resultsofpreviousrepeateddosetoxicitystudies;theMOAandtheMTD.”

Whileonestudyhasaslightdecreaseinbody-weightgain,therearenoindicationsinanyotherstudiesofanexceedanceindosethatwouldsupportignoringthefindingsfromanyexposuregroup.

EPA[33]usesaslightlydifferentcriteriatodeterminewhichdosetoincludeorexcludebasedonanearlierOECDdocument.ThesearespelledoutinEPA’sguidelinedocumentforcarcinogenicityriskassessment[33]

“Othersignsoftreatment-relatedtoxicityassociatedwithanexcessivehighdosemayinclude(a)significantreductionofbodyweightgain(e.g.,greaterthan10%),(b)significantincreasesinabnormalbehavioralandclinicalsigns,(c)significantchangesinhematologyorclinicalchemistry,(d)saturationofabsorptionanddetoxificationmechanisms,or(e)markedchangesinorganweight,morphology,andhistopathology.Itshouldbenotedthatpracticalupperlimitshavebeenestablishedtoavoidtheuseofexcessivelyhighdosesinlong-termcarcinogenicitystudiesofenvironmentalchemicals(e.g.,5%ofthetestsubstanceinthefeedfordietarystudiesor1g/kgbodyweightfororalgavagestudies[OECD,1981]).”Asbefore,thisappliestoonlyonestudypresentedinthisreview.


50

Bothoftheseguidelinesmakegoodscientificsense.Inthe12acceptablerodentcarcinogenicitystudiesincludedinthisevaluation,nostudyhadsufficienttoxicityatthehighestdosetojustifyremovingthehighestdosefromtheanalysis.Hence,theanalysespresentedheredidnotdropthedoses>1000mg/kg/day.ThisisalsosupportedbyonememberoftheEPA’sSAP[54].

Twentychronicrodentcarcinogenicitystudieshavebeendoneusingglyphosateasthetestcompound.Eightofthesestudiesareunacceptableforuseinanevaluationofcausalityleavingsevenstudiesinratsandfivestudiesinmice.Becauseofthelargenumberofevaluationsdoneinanindividualanimalcarcinogenicitystudy,thereisconcernthatthefalse-positiveratescouldbeexaggerated.Forexample,if20evaluationsaredoneandafindingisdeemedsignificantifpTrend<0.05,thenyouwouldexpectthat20*0.05=1evaluationwouldbepositivesimplyduetochance.

Table15:Observedversusexpectedtumorsiteswithsignificanttrendsinthe12acceptablerodentcarcinogenicitystudiesusingglyphosate.

Species Strain Sex TotalSites1

Exp.<0.05

Obs.<0.05

Tumors2p<0.05 Exp.<0.01

Obs.<0.01

Tumorsp<0.01

Rat(7studies)

Sprague-Dawley

(4studies)

M 86 4.3 4 TICT,TFAC,KA,HA 0.9 2 TICT,KAF 102 5.1 1 TCCC 1.0 1 TCCC

Wistar(3studies)

M 64.5 3.2 2 HA,SK 0.6 1 HAF 76.5 3.8 2 MC,MAC 0.8 1 MAC

Mouse(5studies)

CD-1(4studies)

M 42 2.1 8 KA,KC,KAC,HS(2)3,ML(2),LAC

0.4 5 KA,KC,HS(2),ML

F 60 3 1 H 0.6 1 HAlbino(1study)

M 10.5 0.5 0 0.1 0 F 15 0.8 1 H 0.2 1 H

Rats(7studies)

All(7studies)

M 150.5 7.5 6 TICT,KA,HA(2),TFAC,SK 1.5 3 TICT,KA,HAF 178.5 8.9 3 TCCC,MC,MAC 1.8 2 TCCC,MAC

Both 329 16.5 9 TICT,KA,HA(2),TFAC,SK,TCCC,MC,MAC

3.3 5 TICT,KA,HA,TCCC,MAC

Mice(5studies)

All(5studies)

M 52.5 2.6 8 KA,KC,KAC,HS(2),ML(2),LAC

0.5 5 KA,KC,HS(2),ML

F 75 3.8 2 H(2) 0.7 2 H(2)Both 127.5 6.4 10 KA,KC,KAC,HS(2)3,H(2),

ML(2),LAC1.3 7 KA,KC,HS(2),

H(2),MLAll

(12studies)All

(12studies)M 203 10.1 14 TICT,KA(2),HA(2),TFAC,

SK,KC,KAC,HS(2),ML(2),LAC

2.0 8 TICT,HA,KA(2),KC,HS(2),ML

F 253.5 12.7 5 TCCC,MC,MAC,H(2) 2.5 4 TCCC,MAC,H(2)

Both 456.5 22.8 19 TICT,KA(2),HA(2),TFAC,SK,KC,KAC,HS(2),H,ML(2),LAC,TCCC,MC,

MAC

4.6 12 TICT,HA,KA(2),KC,HS(2),H(2),ML,TCCC,MAC

1NumberofsitesexaminedisbaseduponsuggestionsbyDr.J.HasemaninhiswrittentestimonytotheEPA;malemice–10.5sites;femalemice–15sites;malerats–21.5sites;femalerats–25.5sites2Tumorabbreviationsare:KA–kidneyadenoma;KC–kidneycarcinoma;KAC–kidneyadenomaorcarcinoma;HS–hemangiosarcoma;H–hemangioma;HA–hepatocellularadenoma;LAC–lungadenomaoradenocarcinoma;ML–malignantlymphoma;MC–mammaryglandcarcinoma;MAC–mammaryglandadenomaorcarcinoma;TCCC–thyroidC-cellcarcinoma;TFAC


51

TheEPAaskedtheSAPtocommentonitsevaluationofglyphosate[61]atameetinginWashington,DCinDecember2016[54].Manycommentswerereceivedfromoutsideexpertsatthismeeting;onesuchsetofcommentscamefromDr.J.K.Haseman(2016)[111].Haseman(2016)directlyaddressedthefalse-positiveerrorrateandconcludedthattheresultsseeninthesestudieswereduetochance.Hedidthisbydecidinghowmanyevaluationswerelikelyforeachstudy(brokenintosex-by-speciesgroups)andthenaggregatingthefindings.Heconcludedthattheeffectivenumberofanalyseswere10.5inmalemice,15forfemalemice,21.5formalerats,and25.5forfemalerats.Haseman(2016)madetwoassumptionsinhisanalysisthatarenotvalid.Thefirstwasthatallofthepossibletrendtestshadbeendoneonallofthesitesheconsideredreasonableforsuchanevaluation.Heidentifiedeightpositivefindings.However,EPAhadnotevaluatedallofthesitesnorhadtheyconsidereddoingaformalanalysisusinghistoricalcontroldata.EPAidentifiedeightsex/speciesgroupsthathadatmostonepositivetumorfindingusingthetrendtestwithpTrend≤0.05.InTables1-14above,Ihaveidentified19tumorswithpTrend≤0.05orpHist≤0.05and12withpTrend≤0.01orpHist≤0.01(Table15).Secondly,Dr.HasemanassumedonecouldaggregateallthestudiesintoonelargeanalysisofType-1error.However,inferenceinthesestudiesisalwaysmadebysex/species/strain(e.g.glyphosatecauseshemangiosarcomasinmaleCD-1mice;notglyphosatecausescancerinrodents),andtheanalysisshouldhavebeendonebygroupingeachseparately.Table15showstheseanalysesaswellastheaggregatedanalysisforalloftheacceptablestudies.

WiththeexceptionofmaleSprague-Dawleyrats,theobservednumberoftumorsareatorneartheexpectednumberforthedifferentsex/straingroupsinrats(Table15).FormaleSprague-Dawleyrats,0.8caseswithpTrend≤0.01orpHist≤0.01areexpectedandtwowereobserved(p=0.21).InfemaleCD-1miceandSwissAlbinomice,theexpectedandobservednumbersareapproximatelyequal.However,inmaleCD-1mice,therewere2.1tumorsexpectedforpTrend≤0.05orpHist≤0.05andeightwereobserved(p<0.001)andtherewere0.4expectedforpTrend≤0.01orpHist≤0.01andfivewereobserved(p<0.001).Thisclearlycouldnothaveoccurredbychancealone.Evenifoneincorrectlygroupsallsexesandspeciestogether,thereare4.6expectedresponsesforpTrend≤0.01orpHist≤0.01and12observed(p<0.001).Thus,chancedoesnotexplainthepositiveresultsseeninthesestudies.

ConclusionforAnimalCarcinogenicityStudies

Thereareseveralgeneralissuesthatpertaintoallanimalcarcinogenicitystudies.Thereisconsiderablegeneticvariabilityacrossanimalstrainsbothovertimeandspace.Itisdifficulttocompareexperimentsdoneindifferentlaboratoriesevenwhenusingthesamestrainofanimal.ThisisobviouswhenyouexaminetheratesforhepatocellularadenomasinWistarratsacrossthethreestudiesusingthisstrain.Thus,eachstudy

–thyroidfollicularcelladenomaorcarcinoma;TICT–testesinterstitialcelltumor;SK–skinkeratocanthoma3(x):xstudieswiththisresult


52

shouldbeconsideredseparatelywithregardtothefindingsinthatstudybeforebeingcomparedacrossstudies.

Theuseofap-valueof0.05asthecutoffforincreasingtumorincidencedoesnotaccountfortrendsinthedataacrossmultiplestudies.Threestudieswithmarginalresponsesof6-8%inagiventumorcould,whenpooledforanalysis,leadtohighlysignificantfindings.Thisissueiswell-recognizedinepidemiologybutnotusuallyconsideredintoxicologybecauseofalackofreplicatestudies.Thiscaseisfairlyuniquebecauseofthelargernumberofstudiesavailableforanalysisandrequiresamorerigorousevaluationofthedatasuchasthepooledanalysispresentedinthisreport.

Poolingofthedatafortheevaluationofreplicatestudiesmakessenseasitaddressesthequestion“Doesthedataasawholesupportafindingofincreasedcancerincidenceinthesestudies?”Sometoxicologistsmayarguethatthestudiesarenotreplicatesandhencecannotbepooled.Butiftheyarenotreplicates,thentheycannotbecomparedtoseeifthereisconsistencyacrossthestudies.Thisisbecausetheremaybesomesubtlechangefromonestudytoanotherthatleadstoapositivefindinginonestudybutanegativefindinginotherstudies.Thus,eitherthestudiesarenotgoodreplicatessoyoucannotcompareacrossstudiesandyoucannotpoolthem,ortheyaregoodreplicatessoyoucancompareacrossstudiesandyoucanpoolthem.Thereisnoargumentthatwouldsupportacomparisonacrossstudiesthatisappropriatewhenpoolingisinappropriate.

Thereweresevenratstudiesandfivemousestudiesthatwereofsufficientqualityandwithsufficientdetailsavailableforinclusioninthisevaluation.

Glyphosatehasbeendemonstratedtocausecancerintwostrainsofratsandonestrainofmice.GlyphosatecauseshepatocellularadenomasinmaleWistarratsand,toalesserdegree,inmaleSprague-Dawleyrats,mammaryglandadenomasandadenocarcinomasinfemaleWistarrats,skinkeratocanthomasinmaleWistarrats,andkidneyadenomasandthyroidC-celladenomasandcarcinomasinmaleSprague-Dawleyrats.Glyphosatecauseshemangiosarcomas,kidneytumorsandmalignantlymphomasinmaleCD-1miceandhemangiomasinfemaleCD-1miceandpossiblycausesmalignantlymphomas,kidneyadenomasinmaleSwissalbinomiceandhemangiomasinfemaleSwissalbinomice.Thus,glyphosatecausescancerinmammals.

MechanismsRelatingtoCarcinogenicity

Manyhumancarcinogensactviaavarietyofmechanismscausingvariousbiologicalchanges,takingcellsthroughmultiplestagesfromfunctioningnormallytobecominginvasivewithlittleornogrowthcontrol(carcinogenic).HanahanandWeinberg(2011)[112]identifiedmorphologicalchangesincellsastheyprogressthoughthismultistageprocessandcorrelatedthesewithgeneticalterationstodevelopwhattheyrefertoasthe“hallmarksofcancer.”Thesehallmarksdealwiththeentireprocessofcarcinogenesisandnotnecessarilywiththereasonsthatcellsbeginthisprocessortheearlystagesintheprocesswherenormalprotectivesystemswithinthecellsremove


53

potentiallycancerouscellsfromthebody.Whiletumorsthatarisefromachemicalinsulttothecellmaybedistinctfromothertumorsbymutationalanalysis,theyallexhibitthehallmarksasdescribedbyHanahanandWeinberg(2011).

Systematicreviewofalldataonthemechanismsbywhichachemicalcausescanceriscomplicatedbytheabsenceofwidelyacceptedmethodsforevaluatingmechanisticdatatoarriveatanobjectiveconclusiononhumanhazardsassociatedwithcarcinogenesis.Suchsystematicmethodsexistinothercontexts[113],butareonlynowbeingacceptedasameansofevaluatingliteratureintoxicologicalevaluations[114-117].

Inthisportionofthereport,Iamfocusingonthemechanismsthatcancausecancer.Smithetal.(2015)[37]discussedtheuseofsystematicreviewmethodsinidentifyingandusingkeyinformationfromtheliteraturetocharacterizethemechanismsbywhichachemicalcausescancer.Theyidentified10“KeyCharacteristicsofCancer”usefulinfacilitatingasystematicanduniformapproachtoevaluatingmechanisticdatarelevanttocarcinogens.These10characteristicsarepresentedinTable16(copiedfromTable1ofSmithetal.(2015)[37]).Whilethereislimitedevidenceonglyphosateformostofthekeycharacteristics,genotoxicity(characteristictwo)andoxidativestress(characteristicfive)havesufficientevidencetowarrantafullreview.

GenotoxicityGenotoxicityreferstotheabilityofanagent(chemicalorotherwise)todamagethegeneticmaterialwithinacell,thusincreasingtherisksforamutation.Genotoxicsubstancesinteractwiththegeneticmaterial,includingDNAsequenceandstructure,todamagecells.DNAdamagecanoccurinseveraldifferentways,includingsingle-anddouble-strandbreaks,cross-linksbetweenDNAbasesandproteins,formationofmicronucleiandchemicaladditionstotheDNA.

JustbecauseachemicalcandamageDNAdoesnotmeanitwillcausemutations.So,whileallchemicalsthatcausemutationsaregenotoxic,allgenotoxicchemicalsarenotnecessarilymutagens.Doesthatmeanthatthegenotoxicityofachemicalcanbeignoredifallassaysusedforidentifyingmutationsincellsfollowingexposuretoachemicalarenegative?Theanswertothatquestionisnoandistiedtothelimitationsintestsformutagenicity(theabilityofachemicaltocausemutationsinacell).Itisunusualtoseeanevaluationofthesequenceoftheentiregenomebeforeexposurewiththesamesequenceafterexposuretodetermineifthegenomehasbeenaltered(mutation).Thereareassaysthatcanevaluateacriticalsetofgenesthathavepreviouslybeenassociatedwithcanceroutcomes(e.g.canceroncogenes),buttheseareseldomapplied.Ingeneral,mutagenicitytestsarelimitedinthenumbersofgenestheyactuallyscreenandthemannerinwhichthesescreenswork.

Becausescreeningformutagenicityislimitedinscope,anygeneticdamagecausedbychemicalsshouldraiseconcernsbecauseofthepossibilityofamutationarisingfromthatgeneticdamage.Inwhatfollows,Iwillsystematicallyreviewthescientificfindingsavailableforevaluatingthegenotoxicpotentialofglyphosate.Thiswillbedividedintosixseparatesourcesofdatabasedonthebiologicalsourceofthatdata:(1)datafromexposedhumans,(2)datafromexposedhumancellsinalaboratorysetting,(3)data


54

fromexposedmammals(non-human),(4)datafromexposedcellsofmammals(non-human)inthelaboratory,(5)datafromnon-mammaliananimalsandothers,and(5)datafromcellsfromnon-mammaliananimalsandothers.Thesesixareasarebasedupontheprioritiesonewouldapplytothedataintermsofimpacts.Seeinggenotoxicityinhumansismoreimportantthanseeinggenotoxicityinothermammals,whichismoreimportantthanseeinggenotoxicityinnon-mammaliansystems.Inaddition,seeinggenotoxicityinwhole,livingorganisms(invivo)carriesgreaterweightthanseeingresponsesincellsinthelaboratory(invitro).Basically,thecloserthefindingsaretoreal,livinghumanbeings,themoreweighttheyshouldbegiven.

Table16:Keycharacteristicsofcarcinogens,Smithetal.(2016)[37]Characteristic Examplesofrelevantevidence1.Iselectrophilicorcanbemetabolicallyactivated

Parentcompoundormetabolitewithanelectrophilicstructure(e.g.,epoxide,quinone),formationofDNAandproteinadducts

2.Isgenotoxic DNAdamage(DNAstrandbreaks,DNA–proteincross-links,unscheduledDNAsynthesis),intercalation,genemutations,cytogeneticchanges(e.g.,chromosomeaberrations,micronuclei)

3.AltersDNArepairorcausesgenomicinstability

AlterationsofDNAreplicationorrepair(e.g.,topoisomeraseII,base-excisionordouble-strandbreakrepair)

4.Inducesepigeneticalterations

DNAmethylation,histonemodification,microRNAexpression

5.Inducesoxidativestress Oxygenradicals,oxidativestress,oxidativedamagetomacromolecules(e.g.,DNA,lipids)

6.Induceschronicinflammation

Elevatedwhitebloodcells,myeloperoxidaseactivity,alteredcytokineand/orchemokineproduction

7.Isimmunosuppressive Decreasedimmunosurveillance,immunesystemdysfunction

8.Modulatesreceptor-mediatedeffects

Receptorin/activation(e.g.,ER,PPAR,AhR)ormodulationofendogenousligands(includinghormones)

9.Causesimmortalization Inhibitionofsenescence,celltransformation10.Alterscellproliferation,celldeathornutrientsupply

Increasedproliferation,decreasedapoptosis,changesingrowthfactors,energeticsandsignalingpathwaysrelatedtocellularreplicationorcellcyclecontrol,angiogenesis

Abbreviations:AhR,arylhydrocarbonreceptor;ER,estrogenreceptor;PPAR,peroxisomeproliferator–activatedreceptor.Anyofthe10characteristicsinthistablecouldinteractwithanyother(e.g.,oxidativestress,DNAdamage,andchronicinflammation),whichwhencombinedprovidesstrongerevidenceforacancermechanismthanwouldoxidativestressalone.


55

Thedatabeingincludedinthisreviewcomefromthepeer-reviewedscientificliterature,thesummariesofreportsinregulatorydocumentsthatareproprietaryandforwhichIhavelimitedaccesstotheoriginalwork,andreportsfromindustrythatareproprietarytowhichIhavebeengivengreateraccess.Allofthesestudiesareincludedintheoverallevaluationofcausation.

GenotoxicityinHumansin-vivo

Threestudieshaveevaluatedthepotentialgenotoxicityofglyphosateformulationsinexposedhumans.Paz-y-Miñoetal.(2007)[118]analyzedthebloodof24exposedindividuals(livingwithin3kilometersofspraying)and21unexposedindividuals(living80kilometersawayfromthesprayingarea)forDNAdamageusingthecometassay.AllstudysubjectswerefromEcuadorandnoneofthecontrolsorexposedindividualssmoked,drankalcohol,tooknon-prescriptiondrugsorhadbeenexposedtopesticidesduringthecourseoftheirnormaldailylives.Exposedandcontrolindividualsdidsomecultivatingandharvestingbutwithoutpesticidesorherbicides.Exposedindividualswereanalyzedwithintwomonthsofsprayingfortheeradicationofplantsassociatedwithillegalnarcotics.Anaverageof200cellsperpersonwererankedbetween0-400dependingontheamountofDNAinthecomet’stailinordertocalculatethemeanamountofDNAdamage.Therewasasignificantdifferencebetweenthemeantotalmigrationlevelofexposedindividualstocontrols(p<0.001).DatawasgivenforeachindividualclassifiedintofivegroupsbasedupontheamountofDNAinthecomet’stail.TherewasclearlyashiftinthedistributionofDNAincellswiththecontrolsneverseeingscoresinthetoptwocategorieswhileallbutthreeexposedhadsomescoresinthetoptwocategories.Inessence,someoftheDNAhadbeenfragmentedbytheexposure.

Inasecondstudybythesamegroup,Paz-y-Miñoetal.(2011)[119]evaluatedthekaryotypes(thechromosomecountoftheindividualsandanyalterationstothechromosomesasseenunderamicroscope)of92peoplelivingin10communitiesinnorthernEcuador.Controlswerefromareaswithoutsprayingandbothcontrolsandexposedsubjectshadnohistoryofexposuretosmokingorothergenotoxiccompounds.Thisstudysawnochangesbetweencontrolsandexposedsubjectsfor182karyotypesevaluated.

Bolognesietal.(2009)[120]studiedwomenofreproductiveageandtheirspousesinfiveareasofColombia,fourofwhicharesubjecttosprayingforeithernarcoticscontrolorsugarcanegrowing.Therewere60subjectsfromtheSantaMartaarea(organiccoffeeisgrownwithouttheuseofpesticides),52fromBoyaca(manualsprayingforillicitdrugs),58fromPutumayo(aerialsprayingforillicitdrugsusingaglyphosateformulation),63fromNariño(sameexposureasPutumayo)and28fromValledelCauca(aerialsprayingofRoundup747(74.7%glyphosate)withoutadditionaladjuvantforsugarcanematuration).Allsubjectswereinterviewedwithastandardizedquestionnairedesignedtoobtaininformationaboutcurrenthealthstatus,healthhistory,lifestyleandpotentialexposuretopossibleconfoundingfactors(smoking,useofmedicinalproducts,severeinfectionsorviraldiseasesduringthelastsixmonths,recentvaccinations,presenceofknownindoor/outdoorpollutants,exposuretodiagnosticx-rays,andpreviousradio-orchemotherapy).InSantaMarta,bloodsamplesweretaken


56

once,duringtheinitialinterview.InBoyaca,bloodsamplesweretakenattheinitialinterviewand1monthlater.InNariño,PutumayoandValledelCauca,bloodsamplesweretakenattheinitialinterview,withinfivedaysaftersprayingand4monthslater.Inlymphocytes,binucleatedcellswithmicronuclei(BNMN)werelowestinSantaMartaandsimilarinthefourexposedregionspriortoexposure.StatisticallysignificantincreasesinBMNMinNariño,PutumayoandValledelCaucawereseenbetweenfirstandsecondsampling.ThemeanBNMNinNariñoandPutumayowasgreaterinrespondentswhoself-reporteddirectcontactwithsprayedfields,butdifferenceswerenotstatisticallysignificant.MultiplelinearregressiondemonstratedstatisticallysignificantincreasesinBMNMinallfourexposedregionspostexposurewhencomparedtopre-exposureandcontrollingforallothervariables(p<0.001).ThelargesttotalchangeinmeanBMNMvaluespre-exposurecomparedtoimmediatepostexposureoccurredinValledelCaucawheresprayingisdoneusingRoundupwithnoadditionaladjuvant.

Kier(2015)[121]identified16additionalstudiesofpesticideusethatincludedsomeexposuretoglyphosate.Elevenofthe16studiesdemonstratedsomedegreeofgenotoxicityinthehumanpopulationsstudiedbutdidnotadequatelyattributetheexposureprimarilytoglyphosatesotheyarenotincludedinthisreview.

Insummary,twoofthethreestudiesinwhichgenotoxicityendpointswereevaluatedinhumansinareaswithexposuretoglyphosatesprayingshowedstatisticallyincreasedchangesinDNAdamageinblood.Inthestrongeststudy,inthreeareaswherechromosomaldamage(micronuclei)wasexaminedinindividualspre-andpost-spraying(<5days)showedstatisticallysignificantincreases.Inoneotherareawherepost-exposuredamagewasmeasuredonemonthafterexposure,therewaslittlechange.

GenotoxicityinHumanCells(invitro)

Studieshaveexploredtheinvitrogenotoxicityofglyphosateusingavarietyofdifferentcelltypes(lymphocytes,fibroblasts,andimmortalizedcellsfromcancersofthelarynx,mouth,bloodandliver)usingseveraldifferentassaysformarkersofgenotoxicitywithorwithoutmetabolicactivation.

Mladinicetal.(2009)[122]inducedDNAstrandbreaks(cometassay)fromexposuretoglyphosate(puritynotgiven)inlymphocytesfromthreehealthyhumandonors(questionnaireusedtoexcludegenotoxicexposures)atconcentrationsof3.5,92.8and580µg/mlwithS9activationandsaweffectsatonlythehighestdosesforcellswithoutS9activation.

Alvarez-Moyaetal.(2014)[123]conductedasimilarstudyusinglymphocytesfromhumanvolunteers(questionnaireusedtoexcludegenotoxicexposures)andexposuretoglyphosate(96%purity)atconcentrationsof0.12,1.2,12and120µg/ml.AsignificantincreaseinDNAstrandbreaks(cometassay)wasseenforallexposuregroupswithacleardose-responserelationshipwithoutmetabolicactivation(metabolicactivationwasnottested).

UsinghumanHEP-2cells,Manasetal.(2009)[124]inducedDNAdamage(cometassay)by


57

glyphosate(96%pure)atallconcentrationsrangingfrom676µg/mlto1270µg/ml(noS9activationtested).Cellviabilityatthehighestconcentrationwasbelow80%andvaluesattheotherconcentrationswerenotgiven.

Monroyetal.(2005)[125]inducedsignificantDNAdamage(cometassay)infibroblastGM38cellsatconcentrationsofglyphosate(technicalgrade,puritynotgiven)rangingfrom676µg/mlto1000µg/mlwithacleardose-responsepattern.Overthissameconcentrationrange,theyalsosawconcentration-dependentdecreasesincellviabilityatalldosesmakingthecometassayresultsdifficulttointerpret.Inasimilaranalysisinthesamepaper,usingfibrosarcomaHT1080cells,theyalsosawconcentration-dependentDNAdamageandlossofcellviability.ActivationbyS9wasnotusedineitherexperiment.

Luekenetal.(2004)[126]inducedDNAdamage(cometassay)infibroblastsGM5757ataconcentrationofglyphosate(98.4%purity)of12,680µg/mlincombinationwithexposureto40or50mMH2O2.ActivationbyS9wasnotusedinthisexperiment.Accordingtotheauthors,cellviabilityatthisexposurelevelwasabove80%.

Kolleretal.(2012)[127]significantlyinducedDNAdamage(cometassay)inhumanTR146cells(buccalcarcinomacells)fromexposuretoglyphosate(>95%purity)inadose-dependentfashionatconcentrationsof20and40µg/ml.Above40µg/ml,therewasasignificantincreaseintailintensityrelativetocontrols,buttheactualamountincreaseddidnotchangeasthedoseincreased(plateau).UsingRoundup(UltraMax)theauthorssawvirtuallythesamelevelofDNAdamageat20and40µg/ml,buttheconcentrationresponsecontinuedtoincreaseabovethatexposure.TheseexperimentsdidnotuseS9activation.TheyalsousedtheCBMNassayinthesamesystemtoevaluatethetotalnumberofmicronucleiinbinucleatedcells(MNI),thenumberofbinucleatedcellswithmicronuclei(BN-MNI),thenumberofnuclearbuds(NB)andthenumberofnucleoplasmicbridges(NPB)causedbyglyphosateandRoundupexposure.Twoendpoints(NB,NPB)hadsignificantincreasesatconcentrationsof10,15and20µg/mlandtwo(MNi,BN-MNi)weresignificantlyelevatedforconcentrationsof15and20µg/ml.EquivalentRoundupexposuresresultedinsignificantincreasesinallfourmeasuresofDNAdamageat10,15and20µg/ml.TheresultsfortheRoundupweregreaterthanforglyphosatealone.

Gasnieratal.(2009)[128]exposedcellsfromthehepatomacelllineHepG2toglyphosate(puritynotgiven)andfourglyphosateformulations.OnlyoneglyphosateformulationwastestedforDNAdamage(cometassay)andtheysawsignificanteffectsatequivalentconcentrationsof0.05µ/mlto4µg/mlofglyphosate(p-valuesnotgiven).Nop-valuesareprovidedandpresentationoftheresultsdoesnotprovideaclearmeanstocomparetheseresultswithotherstudies.Thisstudywillnotbeusedintheevaluation.

Manasetal.(2009)[124]obtainedhumanbloodsamplesfromthreehealthy,non-smokingwomenandthreehealthymenwithnohistoryofpesticideexposure.Lymphocyteswereculturedwithglyphosate(96%purity)atconcentrationsof34,203,and1015µg/mlwithnostatisticallysignificantchangesinchromatidbreaks,


58

chromosomebreaks,chromatidgaps,chromosomegaps,dicentrics,acentricfragments,orendoreduplication.

Mladinicetal.(2009)[129]usedbloodfromthreenon-smoking,healthyvolunteerstoevaluatetheformationofmicronuclei,nuclearbudsandnucleoplasmicbridgesasafunctionofexposuretoglyphosate(98%purity).Significantchangesinmicronucleiwereseenfollowingexposuretoglyphosateat92.8and580µg/mlinS9activatedcells,butnotthosewithoutmetabolicactivation.Changesinnuclearbudswereseenat580µg/mlforbothS9activatedandnon-activatedcellswhilesignificantchangesinnucleoplasmicbridgeswereseenonlyat580µg/mlinS9activatedcells.Thisstudycontainedapositivecontrol(ethylmethanesulfonateat200µg/ml)whichwasalsonegativeinallassays,manytimesshowingeffectsbelowthatseenforglyphosate.

Bolognesietal.(1997)[130]obtainedbloodfromtwohealthyfemaledonorsandexposedittoglyphosate(99.9%purity)oraRoundupformulation(30.4%glyphosate).Atconcentrationsof1000,3000and6000µg/mlofglyphosateandat100and330µg/mlofglyphosateformulation,significantchangesinsisterchromatidexchanges(SCEs)wereseen.At330µg/ml,anon-significantincreaseinSCEswasseenforglyphosatealonethatwasapproximately20%belowthatseenforanequivalentglyphosateexposurefromtheRoundupformulation.ThisstudydidnotconsiderS9activation.

Lioietal.(1998)[124,131]obtainedbloodfromthreehealthydonorsandexposedittoglyphosate(>98%purity).Atconcentrationsof1.4,2.9,and8.7µg/mlofglyphosate,significantchangesinsisterchromatidexchanges(SCEs)andchromosomalaberrationswereseen.ThisstudydidnotconsiderS9activation.

VigfussonandVyse(1980)[132]exposedculturedhumanlymphocytesfromtwopeopletoRoundup(%glyphosateunknown)atconcentrationsof250,2500and25000µg/ml.Resultsforthehighestconcentrationwerenotprovidedduetolackofcellgrowthinculture.SCEswereshowntobesignificantlyincreasedfortheremainingtwoconcentrationsinonedonorandonlyforthelowestconcentrationintheother.WhiletherelativeSCEcountsseeninthispaperaresimilartothosefromBolognesietal.(1997),theabsolutecountsinthecontrolsareroughlythreetimeshigherinthisstudy.ThisstudydidnotconsiderS9activation.

GenotoxicityinNon-HumanMammals(invivo)

Bolognesietal.(1997)[130]exposedgroupsofthreeSwissCD-1malemicebyIntraperitoneal(IP)injectionwithasingledoseofglyphosate(99.9%purity,300mg/kg)orRoundup(900mg/kg,equivalentto270mg/kgglyphosate).Animalsweresacrificedatfourand24hoursafterinjectionandliversandkidneywereremovedtoobtaincrudenucleifromtheadheringtissues.BothtissuesdemonstratedsignificantincreasesinDNAsingle-strandbreaks(p<0.05)atfourhoursforbothglyphosateandRoundupwithnodiscernabledifferencebetweentheresponses.At24hours,thepresenceofstrandbreakswasreducedandnolongerstatisticallysignificantfromcontrols.

Pelusoetal.(1998)[133]exposedgroupsofsix(controls,lowestdosesofglyphosate-saltandRoundup)orthreeSwissCD-1mice(malesandfemales,specificnumbersnot


59

specified,liverandkidneytissuescombinedforanalysis)totheisopropylammoniumsaltofglyphosateorRoundup(30.4%isopropylammoniumsaltofglyphosate)for24hours.DNAadducts(32P-DNApostlabeling)werenotevidentinmiceexposedtotheglyphosate-saltaloneineitherliverorkidney,butwerepresentinliverandkidneyatalltesteddosesofRoundupshowingadose-responsepattern.

Ranketal.(1993)[134]exposedmaleandfemaleNMRImice(threetofivepersex)toglyphosateisopropylaminesalt(puritynotspecified)andRoundup(480gglyphosateisopropylaminesaltperliter)byintraperitonealinjection.After24or48hours(only24hoursforRoundup),polychromaticerythrocytesfrombonemarrowwereextractedandmicronucleicountedfromasampleof1000cells.Nosignificantincreaseswereseenforanyconcentrationinglyphosate-exposedanimals(100,150and200mg/kg)orRoundup-exposedanimals(133and200mg/kgglyphosateequivalentdose).Thepositivecontrols,whilenotstatisticallysignificant,showedanincreaseinmicronuclei.

Bolognesietal(1997)[130]exposedgroupsofthree,fourorsixmaleSwissCD-1micetoglyphosate(99.9%purity)andRoundup(30.4%glyphosate)byintraperitonealinjectionintwoequaldosesgiven24hoursapart.Aftersixor24hoursfollowingthelastexposure,polychromaticerythrocytesfrombonemarrowwereextractedandmicronucleicountedfromasampleof1000cells.Micegiventwodosesof150mg/kgofglyphosateshowedanon-significantincreaseinmicronucleiat6hoursandasignificantincreaseat24hours.Incontrast,micegiventwodosesof225mg/kgglyphosateequivalentofRoundupshowedasignificantincreaseinmicronucleiatbothsixand24hours.Therelativedifferencesinmeanabsoluteincrease(subtractmeanresponseincontrols)inmicronucleiibetweenglyphosateandRoundupat24hourswas3.6whereastherelativedifferenceinglyphosateequivalentdosewas1.5indicatingagreatereffectoftheglyphosateformulation.

Manasetal.(2009)[124]exposedgroupsofmaleandfemaleBalbCmice(groupsizenotgiven,tissuescombinedforanalysis)toglyphosate(96%purity)byintraperitonealinjectionintwoequaldosesgiven24hoursapart.Twenty-fourhourspostexposure,polychromaticerythrocytesfrombonemarrowwereextractedandmicronucleicountedfromasampleof1000cells.Nosignificantincreaseswereseenatdosesof50mg/kgand100mg/kginglyphosate-exposedanimalsbutasignificantincreasewasseenat400mg/kg.Thepositivecontrolsshowedastatisticallysignificantincreaseinmicronuclei(roughlythreetimesthecontrolrate).

Dimitrovetal.(2006)[135]exposedgroupsofeightmaleC57BLmice(tissuescombinedforanalysis)toRoundup(41%glyphosate)viagavageatadoseof1080mg/kg.At6,24,72,96,or120hourspostexposure,polychromaticerythrocytesfrombonemarrowwereextractedandmicronucleicountedfromasampleof4000cells(500peranimal).Nosignificantincreaseswereseen.Theyalsolookedforchromosomaldamageintheseanimalsandsawnosignificantincreases.Thepositivecontrolsshowedastatisticallysignificantincreaseinmicronuclei.

Prasadetal.(2009)[136]exposedgroupsof15maleSwissCD-1micetoRoundup(30.4%glyphosate)byIPinjectionatdosesof25and50mg/kg.At24,48or72hourspost


60

exposure,polychromaticerythrocytesfrombonemarrowwereextractedandmicronucleicountedfromasampleof2000cellsperanimal,fiveanimalspersacrifice.Micronucleiicountsweresignificantlyincreased(p<0.05)atalldosesatalltimesrelativetocontrols.Inaddition,thenumberofcellswithchromosomalaberrationswassignificantlyincreasedforalldosesatalltimes.ThecontrolrateofmicronucleiwassimilartothatofBolognesietal.(1997),butabout50%greaterresponseforadosethatwasapproximately10timessmaller.

Grisoliaetal.(2002)[137]exposedgroupsofSwissmice(sexandsamplesizenotgiven)toRoundup(480gglyphosateisopropylaminesaltperliter)byIPinjectionatdosesof50,100and200mg/kgRoundupintwodosesseparatedby24hours.At24hourspostexposure,polychromaticerythrocytesfrombonemarrowwereextractedandmicronucleicountedfromasampleof2000cellsperanimal.Micronucleicountswerenotincreasedatanydose.ThisexposureappearstobethesameformulationofRoundupusedinthestudybyRanketal.(1993)whichwasalsonegative.

CoutinhodoNascimentoandGrisolia(2000)[138]exposedgroupsofsixmalemice(strainnotgiven)toRoundup(%glyphosatenotgiven)byIPinjectionatdosesof50,100and200mg/kgintwodosesseparatedby24hours.At24hourspostexposure,polychromaticerythrocytesfrombonemarrowwereextractedandmicronucleicountedfromasampleof1000cellsperanimal.Asignificantincreaseinmicronucleiwereseenatadoseof85mg/kg.Noincreasewasseenat42or170mg/kg.Cavusogluetal.(2011)[139]exposedgroupsofsixSwissalbinomicebyIPinjectionwithasingledoseofglyphosateformulation(RoundupUltraMax,450g/lglyphosate,50mg/kgglyphosateequivalentdose).Animalsweresacrificedatthreedaysafterinjection.Micronucleiinnormochromaticerythrocyteswerecountedfromasampleof1000cellsperanimal.Therewasasignificantincreaseinmicronucleiinerythrocytes(p<0.05).G.bilboaeliminatedtheseeffects.

ChanandMahler(1992)[140]exposedgroupsof10maleandfemaleB6C3F1micetoglyphosate(98.6%purity)infeedatdosesof0,507,1065,2273,4776,and10780mg/kginmalesand0,753,1411,2707,5846,and11977mg/kginfemalesfor13weeks.Atsacrifice,polychromaticerythrocytesfromperipheralbloodwereextractedandmicronucleicountedfromasampleof10,000cells.Nosignificantincreaseswereseenatanyofthetesteddoses.

LiandLong(1988)[141]exposedgroupsof18maleandfemaleSprague-Dawleyratstoglyphosate(98%purity)byIPinjectionatadoseof1000mg/kg.At6,12and24hoursposttreatment,6animalsofeachsexweresacrificedandpolychromaticerythrocytesfrombonemarrowwereextractedandmicronucleicountedfromasampleof50cellsperanimal.Thepercentageofcellswithchromosomalaberrationswasnotincreasedatanytimepointfollowingexposure.

GenotoxicityinNon-HumanMammalianCells(invitro)

LiandLong(1988)[141]incubatedChinesehamsterovarycells(CHO-K1BH4)withglyphosate(98%purity)forthreehoursatconcentrationsof5,10,50and100mg/ml.


61

Cellswerethenplatedusing200cellspersampleintriplicateandincubatedfor8-12days.Colonieswerethencountedandresultsexpressedasmutantfrequency.NopositiveresultswereseeninanyexperimentalgroupwithorwithoutS9activation.Itisnotclearwhythereissuchalargedifferenceintheincubationtimesinthevariousgroupsinthisexperiment,norisitclearwhichgroupsincubatedlonger.Inasecondstudyinthesamepublication,non-inducedprimaryrathepatocytes(Fischer344)wereincubatedwithsevenconcentrationsofglyphosate(12.5ng/mlto125µg/ml)for18-20hours.Nosignificantincreaseswereseenfornetgrainspernucleusatanyexposureconcentration.Therewasafour-foldincreaseinthelowestexposuregroupsrelativetocontrolsandtheneveryothertreatedgroupwasbelowthecontrolresponse.Thisisaveryunusualfindingandcouldbeduetothewayinwhichthedataisadjustedfornetgrainsincytoplasm.Theauthorscalculatednetgrainspernucleusbysubtractingthehighestcytoplasmiccountfromthenuclearcount;ifcytoplasmiccountisincreasedbyglyphosatethiscouldbiasthefindingsmakinganyincreaseinnuclearcountdisappear.Nodataisprovidedtoresolvethisissue.

Roustanetal.(2014)[142]incubatedChinesehamsterovarycells(CHO-K1)withglyphosate(puritynotprovided)forthreehoursatconcentrationsof2,5,10,15,17.5,20,and22.5mg/ml.Cellswerethenplatedusing200cellspersampleintriplicateandincubatedfor24hours.Foreachexposureconcentration,2000bi-nucleatedcellswereexaminedformicronuclei.NopositiveresultswereseeninanyexperimentalgroupwithoutS9activationbutthefourhighestexposuregroupsweresignificantwithaclearconcentration-responsepatternwhenS9activationwaspresent.

Lioietal.(1998)[131]exposedlymphocytesfromthreeunrelatedhealthycowstoglyphosate(>98%purity)for72hourstoconcentrationsof3,14.4and28.7µg/mlwithoutS9activation.Chromosomalaberrationsscoredfrom150cellsweresignificantlyincreased(P<0.05)forallexposureconcentrationsofglyphosatewithaclearconcentration-responsepattern.Similarly,SCEspercellwereincreasedatallconcentrations(p<0.05)butnoconcentrationresponsepatternwasevident.

SivikovaandDianovsky(2006)[143]exposedlymphocytesfromtwohealthyyoungbovinebullstoglyphosateformulation(62%glyphosate)for2,24and48hoursusingconcentrationsof4.7,9.5,23.6,47.3,94.6and190µg/mlwithoutS9activation.Chromosomalaberrationsscoredfrom100cellswerenotsignificantlyincreased(P<0.05)withoutS9activationforany24-hourexposureconcentrationofglyphosate(2-and48-hoursexposureswerenotdone).SCEspercellwereincreasedatall24-hourexposureconcentrations(p<0.05)exceptthelowestconcentration.At48-hours,significantincreasesofSCEspercellwereseenatconcentrationsatorabove47.3µg/ml(2-hourexposureswerenotdone).Finally,aftertwohoursofexposurewithS9activation,significanteffectswereseenat5and10µg/mlbutnotat15µg/ml(24-and48-hourexposureswerenotdoneforS9activation).

Holeckova(2006)[144]exposedlymphocytesfromtwohealthyyoungbovinebullstoglyphosateformulation(62%glyphosate)for24hourstoconcentrationsrangingfrom28to1120µmol/LwithoutS9activation.Asignificantincreaseinpolyploidywasobservedat56µmol/L,allothercomparisonswerewithoutsignificance.However,this


62

onefindingcannotbeeasilydismissedbecauseallexposuregroupsabovethisconcentrationhadtoofewcellsforevaluation.ThisstudydidnotconsiderS9activation.

GenotoxicityinNon-HumanSystems(invivoandinvitro)

Fourstudies[123,145-147]infishhaveseenpositiveresultsforgenotoxicity(DNAstrandbreaks,differentassays)followingexposuretoglyphosate.Inaddition,onestudy[148]inoysterspermandembryosexposedtoglyphosatesawnoincreaseinDNAdamage(cometassay)andonestudy[149]intwostrainsofDrosophilamelanogastershowedanincreaseinmutations(wingspottest)atthehigherdosesofexposure.

Fourteenstudies[137,145,147,150-160]inmultiplefishspeciesevaluatedtherelationshipbetweenvariousglyphosateformulationsandgenotoxicitywithallstudiesshowingpositiveresultsforvariousendpoints(DNAstrandbreaks,micronucleusformation,andchromosomalaberrations).Twoofthestudies[150,152]werenegativeformicronucleusformationafterexposuretoglyphosateformulationsandoneofthese[150]wasalsonegativeforchromosomalaberrationsbutbothwerepositiveinothermarkersofgenotoxicity.Twostudies[161,162]demonstratedgenotoxicity(DNAstrandbreaks,micronuclei)incaimanfromin-vivoexposuretoaglyphosateformulation.Threestudies[163-165]demonstratedgenotoxicity(DNAstrandbreaks,micronucleusformation)infrogsortadpolesfromexposuretoglyphosateformulations.Onestudy[148]inoysterspermandembryos,onestudy[166]inclamsandonestudy[167]inmusselsexposedtoaglyphosateformulationsawnoincreaseinDNAdamage(cometassay).Onestudy[168]insnailssawincreasedDNAdamage(cometassay)followingexposuretoaglyphosateformulation.Twostudies[169,170]inwormssawmixedresultsforDNAdamage(cometassay)withoneofthesestudies[169]showingapositiveresultformicronucleusformation.Onestudy[171]inDrosophilamelanogastershowedanincreaseinsex-linkedrecessivelethalmutations.

Inthepublishedliterature,fivestudiesevaluatedtheimpactofglyphosateininvitrosystems.Twoofthesestudies[172,173]lookedatgenotoxicityofglyphosateincombinationwithUVBradiationandsawsignificantincreasesinDNAstrandbreaks(FADUassay)inbacteriawithoutmetabolicactivation.Onestudy[174]ineukaryotefishsawasignificantincreaseinDNAstrandbreaks(cometassay)withoutS9activation.Anotherstudy[141]showednoincreaseinreversemutationsintwostrainsofbacteriawithandwithoutS9activation.

Williamsetal.(2000)[175]summarizedtheliteratureregardingtheuseofreversemutationassaysinS.typhimurium(AmesTest).Fourstudiesusingglyphosateandfivestudiesofglyphosateformulationswereallnegative.Theycitedonestudy[134]ofaglyphosateformulationthatwaspositivewithS9activationandnegativewithoutS9activation.However,thisstudywaspositivewithS9activationinTA100cells,negativewithS9activationinTA98cells,negativewithoutS9activationforTA100cellsandpositivewithoutactivationforTA98cells.Theyalsosummarizedtwostudiesofglyphosateine.colithatwerenegativewithandwithoutactivation.

Twoadditionalstudies[141,176]ofglyphosateusingreversemutationassaysareavailable


63

fromthescientificliterature,bothofwhicharenegative.

RegulatoryStudies

EFSA[89]cited14reversemutationassaysinS.typhimurium(AmesTest),mostofwhichweretestedinstrainsTA98,100,1535,1537(TableB.6.4-1).All14studiesarelistedasnegativebyEFSA.Actualdataisprovidedforonlyoneofthe14studiesandthisstudyisclearlynegative.EPA[61]cited27reversemutationassaysinS.typhimurium(AmesTest),mostofwhichweretestedinstrainsTA98,100,1535,1537(EPATable5.1).All27studiesarelistedasnegative.Nodataisprovidedforanyofthestudies.KierandKirkland(2013)[177]citedresultsfrom18bacterialreversemutationassaysofglyphosateand16ofglyphosateformulations.Tabulatedresultsandbackgroundinformationwereprovidedforall34studies.Sixstudiesofglyphosatealonedemonstratedpositivefindingsinoneormoregroups.

EFSA[89]citesthreestudiesofgenemutationsinmammaliancells,allofwhicharelistedasnegative(EFSATableB.6.4-5),twousethemouselymphomaassay,andoneusestheChinesehamsterovarycell/hypoxanthine-guaninephosphoribosyltransferase(CHO/HGPRT)mutationassay.EPA[61]citesfourstudies,threeofwhichappeartobethesameasthosecitedbyEFSA(EPATable5.2)andthefourthisanothermouselymphomaassay.Allfourarelistedasnegative.KierandKirkland(2013)[177]citetwoofthemouselymphomastudiesandprovidetabulateddata.NeitherstudyshowsanyindicationofastatisticallysignificantincreaseinmutationfrequencyatthethymidinekinaselocusofL5178mouselymphomatk(+/-)cells.EFSA[89]citesoneinvitrostudyofDNAdamageandrepairinmammaliancellswhichislistedasnegative(EFSATableB.6.4-6).ThisstudyisofunscheduledDNAsynthesis(UDSassay)inprimaryratlymphocytes.Theyalsolistfivestudiesofchromosomeaberrations(EFSATableB.6.4-8),whicharecharacterizedasnegative.TwostudiesareinhumanlymphocytesandtwoareinChinesehamsterlung(CHL)cells.DataforoneofthestudiesinCHLisprovidedintabularformandisclearlynegative.EPA[61]citeseightinvitrostudiesofchromosomeaberrationsinmammaliancells(EPATable5.3);twoofthesestudiesmatchstudiesintheEFSAreport.Fourofthestudiesarefromtheliterature[124,131,143,178]andarereviewedabove.Surprisingly,EPAreferstothestudybyManasetal.(2009)[124]asnegativealthoughitwasclearlypositiveinthecometassay.,Additionally,EPAreferstothestudybySivikovaandDainovsky(2006)[143]asnegativeeventhoughtheysawcleareffectsofglyphosateonSCEs.Basically,allfouroftheliteraturestudiescitedbyEPAarepositiveyetEPAlistsonlytwoofthefouraspositive.Theremainingfourstudiesarenotedasnegative;however,nodataissuppliedforthesestudies.KierandKirkland(2013)[177]citeseightliteraturestudies(allreviewedabove)andthreeregulatorystudieswithglyphosateexposure.Thethreeregulatorystudiesarelistedasnegative,andthedataareavailableasatableinthesupplementmaterialtoKierandKirkland(2013);thesestudiesarenegativeatalltestedconcentrationsinCHLcells;onematchesthestudydataprovidedbyEFSA[89].


64

EFSA[89]citesninemicronucleusassays,threeinSwissAlbinomice,twoinNMRImice,twoinCD-1mice,oneinSprague-Dawleyrats,andoneinCDrats(EFSATableB.6.4-12).TheylistonestudyinSwissAlbinomiceasweaklypositiveinmales,onestudyinCD-1miceaspositiveatthehighestdose(dataforthisstudyisprovided)andallotherstudiesasnegative.TheydiscardonestudywithlowdosesinmaleSwissmice,butthetablesprovidedforthisstudyshowaclearlysignificantresultatthehighestdoseused(30mg/kg)andcleardose-response.Theyprovidedatafortwoofthenegativestudieswhichindicatethesestudieswereindeednegative.EPA[61](EPATable5.5)cites20micronucleusassays,fourareavailableinthescientificliteratureandthreearereviewedabove(thefourthreference[179]wasunavailabletomeatthetimeofpreparationofthisreport).Theremaining16studiesincludesixstudiesinSwissAlbinomice,fourstudiesinCD-1mice,threestudiesinNMRImice,twostudiesinSprague-DawleyratsandonestudyinWistarrats.SinceEFSAdoesnotprovidenamesassociatedwiththeirmicronucleusstudies,IcannotdetermineifanyofthestudiescitedbytheEPAarethesameasthosecitedbyEFSA.EPAliststwooftheliteraturestudiesaspositiveandtwoasnegative(matchingmyreviewsforthethreestudiesIhaveaccessto)andallbutoneoftheregulatorystudiesasnegative(theonepositivestudywasinSwiss-Albinomice).KierandKirkland(2013)[177]cite12regulatorymicronucleusassaysofglyphosateandprovidedatatablesforall12.All12ofthesestudiesarecitedbyEPA.KierandKirkland(2013)list11studiesasnegativeandoneasinconclusive.However,fourofthestudiesshowpositiveeffectsinatleastonesex-by-treatmentgroup.Oneofthesefourstudiestheylistasinconclusiveandtheremainingthreestudiesaredeterminedtobenegativebecausetheresponseiswithintherangeofthehistoricalcontrols.Aswasdiscussedfortheanimalcarcinogenicitystudies,thecorrectgrouptouseistheconcurrentcontrol.KierandKirkland(2013)[177]alsocite12regulatorystudiesandthreeliteraturestudieswhereanimalsareexposedtoaglyphosateformulation.Twooftheliteraturestudiesarereviewedaboveandtheremainingstudy[179]wasunavailable.Dataforthe12regulatorystudiesareallprovidedintablesbyKierandKirkland(2013)andshowtwopositivestudiesinCD-1miceandnegativestudiesfortheremaining10.

SummaryforGenotoxicity

Thisisacomplicatedareafromwhichtodrawaconclusionduetothediversityofthestudiesavailable(therearemultiplespecies,multiplestrainswithinaspecies,multiplecelltypesfrommultiplespecies,differinglengthsofexposure,differingtimesofevaluationafterexposure,differingexposures,numerousmarkersofgenotoxicity,andfinallybothglyphosateandmultipledifferentglyphosateformulations).Therearethreestudiesthatevaluatethegenotoxicityofglyphosateinhumansdirectly,36experimentsineightstrainsofmice,threestudiesinrats,ninestudiesinhumanlymphocytesandfourstudiesinotherhumancells,12studiesinnon-humanmammaliancelllines(twousingmousecells,fiveusinghamstercells,twousingratcellsandthreeusingcellsfromcows),alargenumberofstudiesinawidevarietyofnon-mammalianspecies,andaplethoraofstudies,mostlyidentical,inbacteria.

Someconclusionsarestraightforward”;glyphosatedoesnotappeartocausereversemutationsforhistidinesynthesisinSalmonellatyphimurium,regardlessofwhether


65

thesereversemutationsareduetoframeshiftmutationsorpointmutations.Iamcautiousinthisdeterminationbecausetherewereseveralstudieswithpositiveresults,butnoclearpatternisevident.Thereisampleevidencesupportingtheconclusionthatglyphosateformulationsandglyphosatecancausegenotoxicityinnon-mammaliananimalspecies.ThisclearlyindicatesthatbothglyphosateandtheformulationsareabletocauseinjurytoDNA.Sowhilefindingsofgenotoxicityinthesespeciesdonotspeakdirectlytothehazardpotentialinhumans,theydosupportacauseforconcern.

Themoreimportantstudiesarethosethathavebeendoneusingmammaliansystems,humancellsanddirecthumancontact.Table16summarizesthesestudiesinasimpleframeworkthatallowsalloftheexperimentaldatatobeseeninoneglance.Thistabledoesnotaddressthesubtletyneededtointerpretanyonestudy,butsimplydemonstrateswhenastudyproducedpositiveversusnegativeresults.

Clearly,forinvitroevaluationsinhumancells,themajorityofthestudieshaveproducedpositiveresults.Therewasonlyoneregulatorystudyevaluatingglyphosategenotoxicityinhumanlymphocytesfromhealthyvolunteersandthatstudywasnegative.Thestudywasnotsignificantlydifferentfromtheothersixstudiesinthiscategory,fiveofwhichproducedpositiveresults.Themajorityofthesestudiesusedeitherthecometassay(asimplewayformeasuringanytypeofDNAstrandbreak)ormethodsthatcountedspecifictypesofstrandbreaksinthecells(e.g.SCEs,micronuclei,nuclearbudsandnucleoplasmicbridges).Fromtheseassays,wecanconcludethereisDNAdamage.Forglyphosateformulations,thereareonlythreestudiesinhumansinvivo,twoofwhichwerepositive.

Themagnitudeoftheconcentrationsusedinthesestudiescouldpotentiallyleadtofalsepositivesiftheglyphosateiscausingcytotoxicityinthecells.Allsixstudiesusingthecometassaywerepositivewithnostudyshowinganegativeresponsebelow10µg/mlandmixedresultsbelowthatwithpositiveresultsat0.12and3.5µg/mlandnegativeresultsat2.91and10µg/ml.Ingeneral,thecometassaysprovidestrongsupportforgenotoxicity.

Thefourstudiesthatdirectlyaddressedspecifictypesofstrandbreaksincellsfollowingexposuretoglyphosateshowedmarkedlydifferentresponsesacrossthevariousconcentrationsused.Manasetal.(2009)sawnochangesinchromatidbreaks,chromosomebreaks,chromatidgaps,chromosomegaps,dicentrics,acentricfragmentsorendoreduplicationovertherangeofconcentrations3.4-1015µg/ml.Incontrast,Lioietal.(1998)sawchangesinSCEsoverconcentrationsrangingfrom1.4to8.7µg/ml.Bothstudiesweredoneinlymphocytesfromvolunteers.Mladinicetal.(2009)sawsignificantchangesinmicronucleiabove92.8µg/mlandBolognesietal.(1997)sawpositivechangesinSCEsabove1000µg/mlbutnotat330µg/ml.Whilechangeshavebeenseeninthreeofthefourstudies,theactualconcentrationsinwhichthechangesareseenisnotconsistentacrossstudies.IconcludethatglyphosatecausesDNAstrandbreaks,whichisindicativeofgenotoxicity.

Themicronucleusassaysinrodentsexaminingglyphosategenotoxicityareeitherallpositiveinonestrainorallnegativeinonestrainwiththeexceptionofthethreestudies


66

inCD-1miceandfourstudiesinSwissAlbinomice.Forthepositivestudies,wecanaskthequestionofwhether,inthisstrain,theactualnumberofmicronucleiareconsistent.

Table17:Summaryofinvivoandinvitrogenotoxicitystudiesofglyphosateandglyphosateformulationsinmammals1

Invivoorinvitro

Species Celltypeortissue

Glyphosate2 GlyphosateFormulations

NumberPositive

NumberNegative

NumberPositive

NumberNegative

Invivo Humans Peripheralblood

2 1

invitro Humans lymphocytes 5 2(1) 2 Hep2 1 GM38HT1080

1

GM5757 1 TR146 1 1

Invivo SwissCD-1Mouse

Liver/Kidney 1 1 2

Invivo(micro-nucleusassay)

NMRImouse Erythrocytes 4(3) 2(1)SwissCD-1mouse

1 2

BalbCmouse 1 B6C3F1mouse 1 Swissmouse 1(1) 3(2)CD-1mouse 2(2) 1(1) 2(2) 6(6)Swissalbino

mouse1(1) 3(3) 1

C57BLmouse 1Mouse(notspecified)

1

Rats(all) 2(1) 1(1)Invitro Mouse L5178

lymphoma 2(2)

Chinesehamster

Lung 3(3)

Chinesehamster

ovary 1 1

Fischerrat liver 1


67

Rat Lymphocytes 1(1)

Bovine Lymphocytes 1 2

1eachentryinthetablecorrespondstoasinglestudywhereastudyispositiveifatleastonevalidpositivefindingemergedfromthestudyp<0.05;entriesinthetableareonlyforstudieswheredatawasavailabletoreviewincludingdatafromEFSA[89]andKierandKirkland(2000)[177];2numbersarethetotalnumberofstudiesinthiscategory,numbersinparenthesesarethesubsetofstudiesthatareregulatorystudies

InSwissAlbinomice,allfourstudiesweredonewithmalesandfemales.Exposureswerebyoralgavageforthepositivestudy(infemalemice)andIPinjectionbythenegativestudies.Thepositivestudywasat5000mg/kgandthehighestdoseinanyofthenegativestudieswas3024mg/kg.Finally,thecontrolresponseinthepositivestudywas6.7micronucleatedPCEper1000PCEwhereasthecontrolsinthethreenegativestudieswerebetween0and0.6micronucleatedPCEper1000PCE.AnyofthesedifferencescouldeasilyexplainthedifferencesinresponsesothepositiveresultinSwissAlbinomiceshouldbeaccepted.

ForCD-1mice,theonenegativemicronucleusstudywasbyoralgavageinmalesandfemalesatasingledoseof5000mg/kg.Oneofthepositivestudieswasalsobyoralgavageinmalesatasingledoseof2000mg/kg.Becauseofthenatureofstatisticalnoise,thesetwostudiescouldbothoccurwhetherthereisatrueeffectornot.Fortheotherpositivestudy,thedosewasbyIPinjectioninmalemicewithapositiveresponseat600mg/kgthatwasmorethandoubletheresponseofthecontrols.ThesedatasupportthefindingthatglyphosatecancausemicronucleiinmaleCD-1mice,whichisindicativeofgenotoxicity.

Theremaininginvitroassaysinmammaliancellsexposedtoglyphosateshowmixedresults.ThemouselymphomaassayandtheChinesehamsterovaryassaysarelookingforspecificmutationsthatwillallowthesecellstogrowinculture.TheChinesehamsterlung,thetworatassaysandtheassayinbovinelymphocytesaremeasuringDNAdamageandprovidemixedresults.Ingeneral,theseresponsesappeartobenegativewiththeexceptionofthoseseeninbovinelymphocytesthatappeartoshowapositiveincreaseinSCEsfollowingexposuretoglyphosate.

Forglyphosateformulations,themaindifferencebetweenthefindingsforglyphosateandthosefortheglyphosateformulationsisthedirectevidenceforgenotoxicityinhumansandthemicronucleusassaysinSwissmice.TheobservationofgenotoxicityinhumansfollowingexposuretoglyphosateformulationsmustcarrythegreatestweightintheoverallanalysisandtwoofthethreestudieswerepositivewiththestrongeststudybyBolognesietal.(2009)[120]showingthestrongestresponse.

FortheSwissmousestudiesofmicronuclei,thefactthatallthreestudiesarenegativeforglyphosateformulationswhileonestudyispositiveforglyphosatecreatesacleardisagreement.Thepositivestudyisanoralgavagestudywithaneffectseeninmalemiceat30mg/kg/day.Thetwonegativeregulatorystudiesforglyphosateformulationsweredoneat2000mg/kg(about500mg/kgglyphosateequivalent),werealsooral


68

gavagestudiesandwerereplicatesdoneinthesamelaboratoryatdifferenttimes.Theremainingnegativestudyusedglyphosateformulationdosesof50-200mg/kg(25-100mg/kgglyphosateequivalent)butwasdonebyintraperitonealinjection.Withtheexceptionofthedifferentroutesofexposure,thedifferencesbetweenthesestudiescannotberesolved.

Inthiscase,apooledanalysisofthedataisnotpossiblebecauseinalmosteverycase,noonestudyisaclearreplicateofanother.Instead,theappropriateapproachwouldbetodoameta-analysisandevaluatewhichaspectsoftheexperimentaldesignsareimportanttoproducingpositivefindingsofgenotoxicity.Thestudieswiththemostdataforthistypeofanalysisarethevariousinvivoassaysofmicronucleusformation.Ghisietal.(2016)[180]didasystematicsearchtoidentifyallpublishedstudiesevaluatingtheabilityofglyphosateorglyphosateformulationstoinducemicronucleiinvivo.TheauthorsalsousedthedatafromKierandKirkland(2013)[177]summarizedabove.Anexperiment,intheirevaluation,wasdefinedbysex/species/route/formofglyphosatesothatsomestudiesdoingbothsexesusingglyphosateandaglyphosateformulationwillentermultipletimesintotheanalysis.Theyidentified93experimentsfromwhichitwaspossibletodoameta-analysis.Datawereextractedforeachstudyandthelogratioofthemeanofeachexperimentalgrouptothemeancontrolresponse(E+)wasusedtoevaluateeffectsizesinthemeta-analysis.Forthismeta-analyticmean,avaluebelowzerosuggestsnogenotoxicitywhileavalueabovezerosuggestsincreasedgenotoxicity.Atestofheterogeneity(Cochran’sQstatisticdiscussedearlierfortheepidemiologicaldata)wasalsoevaluated.

Figure2isareprintofFigure1fromthestudybyGhisietal.(2016)[180]andisaforestplotfromallstudiestheyevaluatedforglyphosateandglyphosateformulations.ItisclearfromthisplotthatthepredominantresponseispositiveinthesedatawithanoverallgrandmeanresponseacrossallstudiesofE+=1.37anda95%confidenceintervalof(1.356-1.381)(thisishighlystatisticallysignificantwithap<0.0001).TheQtvalueforthegrandmeanwasalsostatisticallysignificantsuggestingthereareotherexplanatoryvariablesinthedatathatwouldhelptoexplaintheoverallvariance.

Categoricalvariableswerethenusedtomakecomparisonsacrossthevariousstratainthedatatoidentifywhichexperimentalconditionsshowthelargestimpactsonthemeanresponse.Mammalianspeciespresentedahighermeaneffect(E+=1.379;1.366-1.391)thannon-mammalianspecies(E+=0.740;0.641-0.840).Glyphosateformulationsshowedagreatermeanresponse(E+=1.388;1.375-1.400)thandidglyphosate(E+=0.121;0.021-0.221),butbothweresignificantlygreaterthanzero.Themeanresponseinstudiesusingonlymaleanimals(E+=1.833;1.819-1.847)wassignificantlydifferentfromzeroaswerestudiesusingbothmalesandfemales(E+=0.674;0.523-0.825)whereasthemeanresponseinstudiesusingonlyfemales(E+0.088;-0.153-0.328)wasnot.Peer-reviewedstudieshadhighermeanresponse(E+=1.394;1.381-1.407)comparedtoregulatorystudies(E+=0.114;0.027-0.202),butbothmeansweresignificantlygreaterthanzero,indicatinganoverallgenotoxiceffect.Othervariableswereexaminedsuchaslengthofexposureandmagnitudeofexposurethathadverylittleimpactontheoverallfindings.


69

Themeta-analysisbyGhisietal.(2016)[180]providesstrongsupportforthehypothesisthatexposuretoglyphosateandglyphosateformulationsincreasestheformationofmicronucleiinvivo.ThismeansthatglyphosateandglyphosateformulationsaredamagingDNAinliving,functioningorganismswithintactDNArepaircapacitystrengtheningthefindingthatglyphosateisgenotoxictohumans.

Figure2:Forestplotofstudiesevaluatingmicronucleusfrequencyinglyphosateexposure,arrangedbyeffectssize.Theplotshowstheestimateoftheresponseratioand95%confidenceinterval(CI)ofeachexperimentincludedinthemeta-analysis.ThenumberbesidethebarsrepresentsthereferencenumberofeachexperimentasinTable1ofGhisietal.(2016)[180].GrandMeanistheoverallmeaneffectssizeofallstudies.[ReprintedfromGhisietal.(2016)[180]]

Fromasimplystatisticalperspective,thereisanotherwayinwhichonecandecideifthepositivefindingsinthemicronucleusassaysinthemiceareduetochance.Fortheglyphosatestudies,ifoneaddsupalloftheindividualexperimentalgroups,thereare79totalgroupswhichcorrespondto79statisticaltests.Assumingthecriticaltestinglevelis0.05forallofthetests,onewouldexpecttoseejustunderfourpositivefindings,yetsixareobserved.Fortheglyphosateformulations,therewere70experimentalgroupssooneexpects3.5positivefindingsyet12areobserved(p<0.01).Overall,therewereatotalof149experimentalgroupsexaminedinmiceformicronucleusformationandweobserved18(7.5expected,p<0.01).Repeatingthisanalysisonthebasisofstudiesinsteadofexperimentalgroups,therewere15studiesforglyphosate(expectednumberis0.75positive)yetsixpositivewereobserved(p<0.01).Fortheglyphosateformulations,therewere18studies(expectednumberis0.9positive)yetsixpositive


70

areobserved(p<0.01).Nowexpandingtoall69studiespresentedinTable17,therewere33positivestudies,buttheexpectationisamere3.5(p<0.01).

Itisclearthatbothglyphosateandglyphosateformulationshavegenotoxicpotential.Butwhichisworse?Ofthe69experimentsinTable17,therewereeightexperimentsfromfiveresearchpublicationsthataddressedbothglyphosateandaglyphosateformulationinthesamelaboratory.Ofthese,twowerenegativeforbothglyphosateandtheformulationanddonotcontributetoadiscussionofrelativepotency.Theremainingsixcanprovidesomeguidanceontherelativepotencyofglyphosatetoglyphosateformulations.InKolleretal.(2007)[127],tailintensityforthecometassaywerevirtuallyidenticalwhentheamountofglyphosateintheformulationwascomparedtotheresultsusingglyphosatealone.Inthesamepaper,micronucleiandrelatedbiomarkerswereconsistentlyhigherintheglyphosateformulationby10-20%.InBolognesietal.(1997),DNAstrandbreaksinliverandkidneyinSwissCD-1micewerevirtuallyidenticalunderequivalentdosesofglyphosateandglyphosateformulations.Intheirmicronucleusassay,theglyphosateformulationwasapproximately50%morepotent.Finally,Bolognesietal.(1997),intheiranalysisofSCEsinhumanlymphocytes,theglyphosateformulationwasapproximatelytwiceaseffectiveasglyphosatealone.InPelusoetal.(1988)[133],DNAadductsinliversandkidneyswereonlyseeninmicetreatedwiththeglyphosateformulation,sothesefindingsarenotlikelytobeduetoglyphosate.Thedatasuggestasmallincreaseinthepotentialforgenotoxicityforglyphosateformulationsrelativetothegenotoxicityonewouldseewithglyphosatealone.

Insummary,thedatasupportaconclusionthatbothglyphosateandglyphosateformulationsaregenotoxic.Thus,thereisareasonablemechanismsupportingtheincreasesintumorscausedbyglyphosateandglyphosateformulationsinhumansandanimals.

OxidativeStress

Oxidativestressreferstoanimbalancebetweentheproductionofreactiveoxygenspecies(freeradicals)inacellandtheantioxidantdefensesthecellhasinplacetopreventthis.Oxidativestresshasbeenlinkedtoboththecausesandconsequencesofseveraldiseases[181-186]includingcancer[37,187-191].Multiplebiomarkersexistforoxidativestress;themostcommonbeingtheincreasedantioxidantenzymeactivity,depletionofglutathioneorincreasesinlipidperoxidation.Inaddition,manystudiesevaluatingoxidativestressusedantioxidantsfollowingexposuretoglyphosatetodemonstratethattheeffectoftheoxidativestresscanbediminished.

OxidativeStressinHumanCells(invitro)

Mladinicetal.(2009)[122]examinedtheinductionofoxidativestressfromexposuretoglyphosate(98%purity)inlymphocytesfromthreehealthyhumandonors(questionnaireswereusedtoexcludeothergenotoxicexposures)atconcentrationsof0.5,2.91,3.5,92.8and580µg/ml.CellswithandwithoutS9activationsawincreasesintotalantioxidantcapacityatonlythehighestdoseforcellswithoutS9activationalthoughaclearconcentrationresponsepatternwasseenwithS9activation.


71

Kwiatkowskaetal.(2014)[192]examinedtheinductionofoxidativestressfromexposuretoglyphosate(puritynotgiven)inerythrocytesobtainedfromhealthydonorsintheBloodBankofLodz,Poland.Erythrocyteswereexposedtoconcentrationsof1.7,8.4,17,42.3,85and845µg/mlandincubatedfor1hour.Oxidativestress(oxidationofdihydrorhodamine123)wassignificantlyincreasedat42.3,85and845µg/lwithaclearconcentration-responsepattern.

Chaufanetal.(2014)[193]examinedtheinductionofoxidativestressfromexposuretoglyphosate(95%purity)andRoundupUltraMax(74.7%glyphosate)inHepG2cells(humanhepatomacellline).Exposureconcentrationswere900µg/mlforglyphosateand40µg/mlfortheglyphosateformulation.Afterincubationfor24hours,oxidativestress(expressedastheactivityofsuperoxidedismutase(SOD),catalase(CAT),glutathione(GSH)andglutathione-S-transferase(GST))wassignificantlyincreased(p<0.0-5)fortheglyphosateformulation(increasedSODactivity)butnotforglyphosatealone.

Coalovaetal.(2014)[194]examinedtheinductionofoxidativestressfromexposuretoaglyphosateformulation(Atanor,48%glyphosate)orwithasurfactant(Impacto)inHep-2cells(humanepithelialcellline).Exposureconcentrationswere376.4µg/mlforAtanor,12.1µg/mlforImpactoand180.2µg/mlforamixtureofthetwo.Afterincubationfor24hours,oxidativestress(measuredasactivityofSOD,CAT,GSH,andGST)wassignificantlyincreasedforImpacto,Atanorandthemixture(CATandGSHonly,p<0.05orp<0.01).

Gehinetal.(2005)[195]examinedtheinductionofoxidativestressfromexposuretoglyphosate(purityunknown)andaglyphosateformulation(Roundup3plus,21%glyphosate)inHaCaTcells(humankeratinocytecellline).Glyphosateinducedcytotoxicityinthecellswhichwasreducedoreliminatedbyantioxidants.Theauthorsattributedthecytotoxicitytooxidativestress.

Elie-Cailleetal.(2010)[196]examinedtheinductionofoxidativestressfromexposuretoglyphosate(purityunknown)inHaCaTcells(humankeratinocytecellline).Exposureconcentrationsrangedfrom1700µg/ltoalmost12,000µg/ml.GlyphosateinducedcytotoxicityinthecellsandincreasedhydrogenperoxideH2O2(dichlorodihydrofluoresceindiacetateassay).Thisstudyusedexceptionallyhighconcentrationsthatmaybeinducingcytotoxicitybymeansthatareindependentoftheoxidativestressobserved.Measuringoxidativestressusingthedichlorodihydrofluoresceindiacetateassayhaslimitations[197,198].

GeorgeandShukla(2013)[199]examinedtheinductionofoxidativestressfromexposuretoaglyphosateformulation(RoundupOriginal,41%glyphosate)inHaCaTcells(humankeratinocytecellline).Exposureconcentrationrangedfrom1.7µg/mlto17,000µg/mlandexposurewasfor24hours.Glyphosatesignificantlyinducedtheformationofreactiveoxygenspecies(dichlorodihydrofluoresceindiacetateassay)atallexposuresinaconcentration-dependentfashion.PriortreatmentofthecellswithN-Acetylcysteinereducedtheimpactofglyphosate,butdidnoteliminateit.Measuringoxidativestressusingdichlorodihydrofluoresceindiacetatehaslimitations[197,198]thataffecttheclear


72

interpretationoftheseresults.

OxidativeStressinNon-HumanMammals(invivo)

Bolognesietal.(1997)[130]exposedgroupsofthreeSwissCD-1malemicebyIPinjectionwithasingledoseofglyphosate(99.9%purity,300mg/kg)orRoundup(900mg/kg,equivalentto270mg/kgglyphosate).Animalsweresacrificedateightand24hoursafterinjectionandliversandkidneywereremovedtoobtaincrudenucleifromtheadheringtissues.Samplesofliverandkidneysfromthesemicewereevaluatedforlevelsof8-hydroxy-2'-deoxyguanosine(8-OHdG)whichisabiomarkerofoxidativestress[200].Therewasasignificantincreaseintheliverof8-OHdGat24hoursfollowingglyphosateexposure,butnotateighthoursandnotinthekidney.Atbotheighthoursand24hours,Roundupincreased8-OHdGinthekidneys,butthemildincreaseseenintheliverat24hourswasnotsignificant.

Cavusogluetal.(2011)[139]exposedgroupsofsixSwissalbinomicebyIPinjectionofaglyphosateformulation(RoundupUltraMax,450g/lglyphosate,50mg/kgformulation).Attheendofdosing,animalswerefastedovernightthensacrificed.TherewasasignificantincreaseinmalondialdehydeinbothliverandkidneyandasignificantdecreaseinGSHinliverandkidneyfromexposuretotheglyphosateformulation.G.bilboaeliminatedtheseeffects.

Jasperetal.(2012)[201]exposedgroupsof10maleand10femaleSwissalbinomiceviaoralgavagefor15daystoaglyphosateformulation(RoundupOriginal,41%glyphosate,50mg/kgglyphosateequivalentdose).Animalsweresacrificedatthreedaysafterinjection.Therewasasignificantincreaseinthiobarbituricacid-reactivesubstances(TBARS)intheliverforbothmaleandfemalemiceatbothdoses(p<0.05).Theconcentrationofnon-proteinthiolswaselevatedinbothdosegroupsformalesandforthehighdoseonlyinfemales(nodose-responsewasseenforthisendpoint).

Astizetal.(2009)[202]exposedgroupsoffourmaleWistarratsbyIPinjectiontoasingledoseofglyphosate(purityunknown,10mg/kg).Animalswereinjectedthreetimesperweekforfiveweeksandthensacrificed.Thiobarbituricacid-reactivesubstances(TBARSassay),proteincarbonyls(PCOSs),totalglutathionelevels,individualglutathionelevels,SODandCATwereallmeasuredasbiomarkersforoxidativestressinplasma,brain,liverandkidney.GlyphosatesignificantlyincreasedTBARSinalltissues(p<0.01),totalglutathioneinbrain(p<0.01),SODinliverandbrain(p<0.01)andCATinbrain.Inafollow-upreport[203],theydemonstratethatlipoicacideliminatesorseverelyreducestheimpactsofglyphosateonthebrain.

Cattanietal.(2014)[204]exposedgroupsoffourpregnantWistarratstoglyphosateformulation(RoundupOriginal,360g/Lglyphosate)indrinkingwaterfromgestationaldays5-15atadoseof71.4mg/kg.Fifteenday-oldpups(2perdam)wereexaminedforoxidativestressmarkersinthehippocampus.PupshadasignificantincreaseinTBARS(p<0.05)andasignificantdecreaseinGSH(p<0.01).

Georgeetal.(2010)[82]exposedgroupsoffourSwissalbinomicetoaglyphosateformulation(RoundupOriginal,36g/Lglyphosate)atadoseof50mg/kg(glyphosate


73

equivalentdose)viaasingletopicalapplication.ProteomicanalysisofskinfromthetreatedanimalssawalterationsinSOD1,CAIIIandPRXII,proteinsknowntoplayaroleinthemanagementofoxidativestress.

OxidativeStressinNon-MammalianSystems

Asforgenotoxicity,oxidativestressfromexposuretoglyphosateandglyphosateformulationshavebeenstudiedinvariousaquaticorganisms;reviewedinSlaninovaetal.(2009)[205].ManyofthestudiesreviewedbySlaninovaetal.(2009)showedassociationswithglyphosateandoxidativestressinvariousorgans.Sincethatreview,additionalstudieshavebeencompletedthatalsodemonstrateapositiveassociationbetweenglyphosateandoxidativestress[147,156-159,206-217].

SummaryforOxidativeStress

Sevenstudiesaddressedoxidativestressinhumancellsandanothersixstudiesaddresseditinmammaliansystems.Inlymphocytesanderythrocytesfromhealthydonors,oxidativestresswasdetectedaslowas580µg/mlinlymphocytesandat42.3µg/mlinerythrocytes.InHep-G2cells,noincreasedoxidativestresswasseenforasingleconcentrationof900µg/l.IntwostudiesinHaCatcells,glyphosateinducedoxidativestressinacontinuousmodelfittotheresultsinonestudyandatthelowestconcentration(1700µg/ml)intheother.Themostconvincingstudiesinhumancellsforoxidativestressarethetwostudiesinhumanblood.

InSwissCD-1malemice,increasedoxidativestresswasseenintheliverat24hours,butnotatfourhoursafterinjectionof300mg/kgglyphosate.Noincreasewasseeninthekidney.InWistarrats,repeatedIPdosingwithglyphosateleadtoincreasedoxidativestressinmultipleorgansusingmultiplebiomarkers.Thus,allofthelaboratorystudiesdemonstratedoxidativestresswithasignificantfindingintheratstudy.

InHep-G2cells,aglyphosateformulationdemonstratedarobustincreaseinoxidativestressat40µg/ml.Giventhenegativeresponseinthiscelllineforglyphosatealone,itmustbeconcludedthatthisresponseisnotduetoglyphosate.InHEP-2cells,aglyphosateformulationdemonstratedarobustincreaseinoxidativestressviamultiplebiomarkersat376µg/mlandwhenasurfactantisadded,at180.2µg/ml.InHaCaTcells,aglyphosateformulationdemonstratedsignificantincreasesinoxidativestressfromdosesstartingaslowas1.7µg/mlinaconcentration-dependentfashion.Nostudieswereavailableinhumanlymphocytes.

InSwissCD-1mice,aglyphosateformulationsignificantlyincreasedoxidativestressinthekidneybutonlydemonstratedamild(non-significant)increaseintheliver.Thisstudyevaluatedoxidativestressattwodifferenttimepointsfollowingexposureandsawresponsesthatdifferedovertime.Thestrongincreaseintheliverforglyphosatebutnotglyphosateformulation,suggestsacomplicatedresponsepatternforpureglyphosateversustheformulationthatcouldbelinkedtothetimesinceexposure.InSwissAlbinomice,aglyphosateformulationdemonstratedincreasedoxidativestressbytwoseparatebiomarkersinboththeliverandthekidney.InasecondstudyinSwissalbinomiceusingadifferentbiomarkerbutasimilardose,increasedoxidativestress


74

wasseeninboththeliverandthekidney.InWistarratpupsexposedinutero,anincreaseinoxidativestresswasseeninthehippocampus.InSwissalbinomice,topicalapplicationofaglyphosateformulationtotheskinresultedinaproteomicfingerprintsuggestingoxidativestresswasincreased.

Thoughtherearefewerstudiesforoxidativestressthanthereareforgenotoxicity,therobustresponseseenhereinhumancellsandinrodentstudiesclearlysupportsaroleforbothglyphosateandglyphosateformulationsininducingoxidativestress.Thus,thereisasecondreasonablemechanismthroughwhichthetumorsseeninhumansandthoseseeninanimalscanbecausedbyglyphosateandglyphosateformulations.

SummaryforBiologicalPlausibility

Intheevaluationofcausality,theevidenceforbiologicalplausibilityisoverwhelming.Glyphosateclearlycausesmultiplecancersinmice,twocancersinthehematopoieticsystemsimilartowhatisseeninhumans,causescancerinrats,isgenotoxicandinducesoxidativestress.Thefindingsareclearforbothglyphosatealoneandforglyphosateformulations.ThereisstrongsupportforbiologicalplausibilityinsupportofacausalassociationofglyphosateandglyphosateformulationswithNHL.

BiologicalGradient

Onlythreeoftheepidemiologicalstudiesprovidedinformationonbiologicalgradientsintheirpublications.

Erikssonetal.(2008)[46]dividedtheircasesandcontrolsintothosewith≤10daysperyearofexposureandthosewith>10daysperyearofexposure.TheORswerecalculatedusingamultivariateanalysisthatincludedagentswithstatisticallysignificantincreasedOR,orwithanOR>1.50andatleast10exposedsubjects.ORsforglyphosatewere1.69(0.70-4.07)for≤10daysperyearand2.36(1.04-5.37)for>10daysperyear.Intheirmultivariateanalysis,latencyperiodsof1-10yearsshowedanORof1.11(0.24-5.08)and>10yearshadanORof2.26(1.16-4.40).Thus,theyshowanincreasewithintensityofexposureandwithlatency.

McDuffieetal.(2001)[50],usingaconditionallogisticregressionanalysiscontrollingformajorchemicalclassesofpesticidesandallothercovariateswithp<0.05,theORfor≤2daysperyearofexposurewas1.0(0.63-1.57)andfor>2daysperyear,theORwas2.12(1.20-3.73).Thus,theyshowanincreasewithintensityofexposure.

DeRoosetal.(2005)[45]usedthreeexposuremetricsintheiranalyses:a)everpersonallymixedorappliedpesticidescontainingglyphosate;b)cumulativeexposuredaysofuseofglyphosate(yearsofusetimesdaysperyear);andc)intensityweightedcumulativeexposuredays(yearsofusetimesdaysperyeartimesintensityofuse).Forexposuremeasurementsbandc,theydividedtherespondentsintotertileschosenaprioritoavoidhavingsparsedatawhendealingwithraretumors.Forcumulativeexposuredaysandusingthelowestexposedtertileasthereferencegroup,theRRsdropwithvaluesof0.7(0.4-1.4)and0.9(0.5-1.6)fortertiles2and3respectivelyadjustedfordemographicandlifestylefactorsandotherpesticides(30,699subjects).When


75

intensity-weightedexposuredaysareexamined,theRRsdropwithvaluesof0.6(0.3-1.1)and0.8(0.5-1.4)fortertiles2and3,respectivelyadjustedfordemographicandlifestylefactorsandotherpesticides(30,699subjects).Thus,theydonotseeabiologicalgradientintheirresponses.However,thehighfrequencyofexposuretomanypesticides(e.g.73.8%wereexposedto2,4-D)meanssubjectswithlowexposuretoglyphosatewerelikelytobeexposedtootheragentsthatmayalsoinduceNHL;thiscouldreducetheRRsinthehigherexposureclassesbecauseitwouldinflatetheRRinthelow-exposurereferentgroup.

Erikssonetal.(2008)[46]andMcDuffieetal.(2001)[50]hadconsistentresultsforintensityofexposureperyear(≤2daysperyear,OR=1.0;≤10daysperyear,OR=1.69;>2daysperyear,OR=2.12;>10daysperyear,OR=2.26).Itisnotpossibletoresolvetheremainingdifferencesbetweenthesethreestudiesnorisiteasytoarguethatonestudyhasmoreweightonthisquestionthananyother.Thestudiesusedifferentmeasuresofexposureortimesinceexposure,aredoneondifferentpopulationsandhavedifferentstatisticalpowertodetectatrend.

Inrodentcarcinogenicitystudies,thereisclearevidenceofabiologicalgradient.

Ingeneral,thereissupportthatabiologicalgradientexistsfortheepidemiologicaldataandthussupportfromthisaspectoftheBradford-Hillevaluation.

TemporalRelationship

Exposuremustcomebeforethecancersoccurotherwisetheepidemiologystudiesareuseless.Inthiscase,itisclearthatexposurecamebeforetheonsetofNHL.TheneedforatemporalrelationshipinthedatasupportingacausalassociationbetweenglyphosateandNHLissatisfied.

Specificity

ThereareothercausesofNHL[218-221]sothisgroupofcancersisnotspecifictoglyphosate.Thereislittlesupportforspecificity.

Coherence

Humans,comingintocontactwithglyphosate,canabsorbthecompoundintotheirbodieswhereithasbeenmeasuredinbloodandinurine[56,222-226].Inlaboratoryanimals,absorption,distributionandeliminationofglyphosateandglyphosatecompoundshavebeenstudied[140,227]andshowthatglyphosategetsintotheanimal’sbodies,distributestonumerousorgansandiseliminatedinurine.Theanimalcancerstudiesclearlydemonstratethatglyphosateinmammalscanhavetoxiceffects.

Mousemodelshavelongservedassurrogatesforhumansinunderstandinganddevelopingtreatmentsformanydiseases.Thesameholdstrueforlymphoidtumorsseeninhumans.Forover30years,mousemodelshavebeenstudiedandevaluatedassurrogatesforNHL[228-232].Thesepublicationsandtheassociatedclassificationsystemsforhumansandmiceindicateacloselinkagebetweenthediseasesinhumansandmice.


76

Thus,coherenceissupportedbytheincreasedriskofmalignantlymphomasinCD-1mice,themarginalincreaseinthesetumorsinSwissmiceandthestrongsimilaritybetweenmalignantlymphomasinmiceandNHLinhumans.

ThereisstrongsupportforcoherenceinthedatasupportingacausalassociationofglyphosateandglyphosateformulationswithNHL.

ExperimentalEvidenceinHumans

Thereisnoexperimentalevidenceinhumanssincepurposelyexposinghumanstoapesticide,especiallyonethatisprobablycarcinogenic,isnotethicalandwouldneverpassreviewbyahumansubject’sadvisoryboard.

Analogy

Iamunawareofanyanalogouscompoundsfromthescientificliterature.This,however,isnotanareawhereIhavesufficientbackgroundtoexpressanopinion.Summary

Table18summarizestheinformationforeachofHill’saspectsofcausality.Forthesedata,causalityisstrengthenedbecausetheavailableepidemiologicalstudiesshowaconsistentpositiveassociationbetweencancerandtheexposure.Thestudiesdonotshowdifferentresponseswithsomestudiesbeingpositiveandothersnegative,nordotheyshowanyheterogeneitywhenanalyzedtogether.And,inanswertoHill’squestion,therelationshipbetweenNHLandglyphosateexposurehasbeenobservedbydifferentpersons,indifferentplaces,circumstances,andtimes.

Causalityisstrengthenedforthesedatabecausethestrengthoftheobservedassociations,whenevaluatedsimultaneously,arestatisticallysignificant,thefindingsareuni-directionalandtheresultsareunlikelytobeduetochance.Eventhoughnoneoftheindividualstudiesproviderelativerisksoroddsratiosthatarelargeandprecise,themeta-analysishasobjectivelyshownthattheobservedassociationacrossthesestudiesissignificantandsupportsapositiveassociationbetweenNHLandglyphosate.

Biologicalplausibilityisstronglysupportedbytheanimalcarcinogenicitydataandthemechanisticdataongenotoxicityandoxidativestress.Whenaddressingbiologicalplausibility,thefirstquestiongenerallyaskedis“Canyoushowthatglyphosatecausescancersinexperimentalanimals?”Inthiscase,theanswertothatquestionisclearlyyes.Glyphosatehasbeendemonstratedtocausecancerintwostrainsofratsandonestrainofmice.Glyphosatehasbeendemonstratedtocausecancerintwostrainsofratsandonestrainofmice.GlyphosatecauseshepatocellularadenomasinmaleWistarratsand,toalesserdegree,inmaleSprague-Dawleyrats,mammaryglandadenomasandadenocarcinomasinfemaleWistarrats,skinkeratocanthomasinmaleWistarrats,andkidneyadenomasandthyroidC-celladenomasandcarcinomasinmaleSprague-Dawleyrats.Glyphosatecauseshemangiosarcomas,kidneytumorsandmalignantlymphomasinmaleCD-1miceandhemangiomasinfemaleCD-1miceandpossibly


77

causesmalignantlymphomas,kidneyadenomasinmaleSwissalbinomiceandhemangiomasinfemaleSwissalbinomice.Thus,glyphosatecausescancerinmammals.Thus,itisbiologicallyplausiblethatglyphosatealonecancausecancerinmammals.

Thenextquestiongenerallyaskedis“Doesthemechanismbywhichglyphosatecausescancerinexperimentalanimalsalsoworkinhumans?”ThebestunderstoodmechanismbywhichchemicalscausecancerinbothhumansandanimalsisthroughdamagingDNAthatleadstomutationsincellsthatthenleadstouncontrolledcellularreplicationandeventuallycancer.Itisabsolutelyclearfromtheavailablescientificdatathatbothglyphosateandglyphosateformulationsaregenotoxic.Thishasbeenamplydemonstratedinhumansthatwereexposedtoglyphosate,inhumancellsinvitro,inexperimentalanimalmodelsandtheircellsinvitroandinvivo,andinwildlife.OnewayinwhichDNAcanbedamagedisthroughthepresenceoffreeoxygenradicalsthatoverwhelmacell’santioxidantdefenses.Glyphosateinducesthistypeofoxidativestress,providingadditionalsupportforabiologicalmechanismthatworksinhumans.

Table18:SummaryconclusionsforHill’snineaspectsofepidemiologicaldataandrelatedscience

Aspect Conclusion Reason

Consistencyoftheobservedassociation Strong

Multiplestudies,allarepositive,meta-analysisshowslittleheterogeneity,differentresearchteams,differentcontinents,differentquestionnaires,noobviousbiasorconfounding

Strengthoftheobservedassociation Strong

Sixcoreepidemiologystudiesallshowthesamemodestincrease,significantmeta-analyses

Biologicalplausibility VeryStrong

Multiplecancersinmultiplespecies,notduetochance,increasedriskofraretumors,convincingevidenceforgenotoxicityandoxidativestress

Biologicalgradient Moderate Clearlyseeninthetwocase-controlstudiesthatevaluatedit,notseeninthecohortstudy

Temporalrelationshipoftheobservedassociation Satisfied Exposureclearlycamebeforecancers

Specificityoftheobservedassociation

Notneeded

NHLhasothercauses,thisdoesnotsubtractfromthecausalargument

Coherence StrongGlyphosateisabsorbed,distributedandexcretedfromthebody,cancersseeninthemicehavestrongsimilaritytohumanNHL

Evidencefromhumanexperimentation Nodata Nostudiesareavailable

Analogy Nodata Nostudiesavailableintheliterature


78

Ingeneral,thereissupportthatabiologicalgradientexistsfortheepidemiologicaldataandthussupportfromthisaspectoftheBradford-Hillevaluation.GlyphosateORsincreasedwithtimesincefirstexposureandwithintensityofuseperyearinthetwocase-controlstudiesthatevaluatedatleastoneoftheseissues.

Thereisclearlythepropertemporalrelationshipwiththeexposurecomingbeforethecancers.

Thehumanevidenceiscoherent.Thebasicfindingsinhumansagreewiththeanimalevidenceforabsorption,distributionandeliminationofglyphosate.Also,oneofthetumorsseeninmicehasalmostthesameetiologyasNHL.

NHLisnotspecifictoglyphosateexposure.ThereisnoexperimentalevidenceinhumansandIdidnotfindanyreferenceswhereresearcherslookedforanalogouscompoundswithsimilartoxicity.

Hill(1965)[36]asks“isthereanyotherwayofexplainingthesetoffactsbeforeus,isthereanyotheranswerequally,ormore,likelythancauseandeffect?”ThereisnobetterwayofexplainingthescientificevidencerelatingglyphosatetoanincreaseinNHLinhumansthancauseandeffect.

Inmyopinion,glyphosateprobablycausesNHLand,giventhehuman,animalandexperimentalevidence,Iassertthat,toareasonabledegreeofscientificcertainty,theprobabilitythatglyphosatecausesNHLishigh.

TheIARCAssessmentofGlyphosate

InMarch2015,theInternationalAgencyforResearchonCancer(anagencyoftheWorldHealthOrganization)broughttogetherseventeenscientists(theWorkingGroup)toevaluatethescientificevidenceonwhetherglyphosatecancausecancerinhumans.Thisgroupalsocontainedoneinvitedspecialist(myself)toaidtheWorkingGroup(WG)ingoingthroughthesciencebutwhowasnotallowedtojoindiscussionsonthefinalconclusionorwriteanypartofthedocument.TheWorkingGroupconcludedthatglyphosatefallsinthecategory“probablycarcinogenictohumans(Group2A)”[56].

TheIARCpreamble[30]guidesWorkingGroupsonhowtoevaluatescientificliteraturetodetermineifsomethingisahazard.AllWorkingGroupsfollowtheseguidelinesandthisprocessisacceptedworldwideasaproperwaytoevaluatetheliteratureforahazard(e.g.,theEuropeanChemicalAgencycitestheIARCreviewprocessasguidanceandthenusestheexactsamewordingasIARCdoestoguidetheirownhazardevaluationprocess[34]).

TheWGexaminedtheepidemiologicaldataandclassifieditas“limitedevidenceofcarcinogenicity,”whichisdefinedtomean“apositiveassociationhasbeenobservedbetweenexposuretotheagentandcancerforwhichacausalinterpretationisconsideredtobecredible,butchance,biasorconfoundingcouldnotberuledoutwith


79

reasonableconfidence.”Thisisapreciseandcleardescriptionofthestrengthoftheevidencefromtheepidemiologicalstudies.

TheWGexaminedtheevidencefromanimalcarcinogenicitystudiesandclassifieditas“sufficientevidenceofcarcinogenicity,”whichIARCdefinesas:“acausalrelationshiphasbeenestablishedbetweentheagentandanincreasedincidenceofmalignantneoplasmsorofanappropriatecombinationofbenignandmalignantneoplasmsin(a)twoormorespeciesofanimalsor(b)twoormoreindependentstudiesinonespeciescarriedoutatdifferenttimesorindifferentlaboratoriesorunderdifferentprotocols.Asinglestudyinonespeciesandsexmightbeconsideredtoprovidesufficientevidenceofcarcinogenicitywhenmalignantneoplasmsoccurtoanunusualdegreewithregardtoincidence,site,typeoftumourorageatonset,orwhentherearestrongfindingsoftumoursatmultiplesites.”BasedonthedataavailabletoIARCatthetimeoftheirreviewandtherestrictionsplacedonthestudiestheycanreviewbythePreamble,thisconclusionisjustifiedandcorrect.

OneofthemajorcriticismsoftheWGreviewwasthattheWGdidnotreviewalloftheanimalcarcinogenicitydatathatwasavailabletotheregulatorybodiesandthuscametothewrongconclusionsontheanimalcancerdata.Inthisreview,Ievaluatedall19animalcarcinogenicityexperimentsthathavebeencollectivelymentionedbyanyagencythatreviewsglyphosate.Wherepossible,Ihaveanalyzedtheoriginaldataandusedsoundstatisticalmethodstotestforsignificantincreasesincancerincidenceinanimalsexposedtoglyphosate.MyconclusionisthattheWGwouldhavecalledthisdata“sufficientevidence”tosupporttheirfindingsdespitenotreviewingtheadditionalstudiesanalyzedherein.Despitethefacttheindustrykeptthesestudiesconfidential,nothingcontainedinthewithheldstudieswouldhavechangedtheWGconclusion.

Onthemechanisticdata,theIARCWorkingGroupreviewedthesamedatathatIreviewed,butIalsoevaluated,wherepossible,theproprietarydatasupportingtheregulatorydecisions.Wherepossible,Ireanalyzedthatdatatobecertaintheresultsbeingpresentedwereaccurate.TheIARCWorkingGroup,usingtheguidelinessetforthintheirPreamble,declaredstrongsupportforthebiologicalmechanismsofgenotoxicityandoxidativestress.AsIhaveshownhere,thereisstrongsupportforthesetwomechanisms,evenwiththeproprietaryevidencefromtheindustrystudies.Thus,theIARCWorkingGroupreachedthecorrectconclusion.

Todecideonafinalclassificationforacompound,theIARCPreambleprovidesguidanceonhowtheclassificationofthethreeareasaretobeused.Ifthedatainhumansis“limited”andthedatafromanimalcarcinogenicitystudiesis“sufficient,”thediscussionsshouldbeginwithClass2A,“theagentisprobablycarcinogenictohumans.”Then,giventheoverallqualityofthedataset,thestrengthoftheevidencefromthemechanisticstudiesandanyadditionalscientificissuesthatneedtobeconsidered,theWorkingGroupwilldeterminewhetherthedatajustifiesadifferentcategory.Inthiscase,theWorkingGroupconcluded2AwastherightcategoryandIstillbelievetheevidencesupportsthatfinding.


80

TheEPAAssessmentofGlyphosate

LikeIARC,theEPAhasguidelinesthataretobefollowedwhenevaluatingscientificliteratureandmakingadeterminationaboutthecarcinogenicpotentialofachemical.ThoseguidelineshavebeendevelopedovermanyyearsandarebasedonsoundscientificguidancethatmyselfandmanyotherscientistshaveprovidedtotheAgency.Fortheirevaluationofglyphosate,theAgencydidnotfollowtheirownguidelines,nordidtheyfollowsoundscientificpractice.ThisopinionisconsistentwiththereviewdonebytheEPAFIFRAScientificAdvisoryPanel[54].Inaddition,theAgencyfailedtofindalloftherelevantanimalcancerstudiesandmisinterpretedseveralofthem.ThemajorproblemswiththeAgencyevaluationare:

• Misinterpretationoftheepidemiologicalevidence,confusingthepotentialforbiasandpotentialforconfoundingwithrealbiasandrealconfounding,allowingthemtogivealmostnoweighttothecase-controlstudiesinfavoroftheonecohortstudy;

• Misinterpretationofthefindingsinthemeta-analysis;

• Failuretoproperlyusehistoricalcontrolsintheanalysisoftheanimalcarcinogenicitystudies;declaringasignificantfindingasnotduetothecompoundifitisintherangeofthehistoricalcontrols;

• Failuretoanalyzealltumorsinallstudiesrelyingupontheindustrysubmissionstohavedonethiscorrectly;

• Failuretofollowtheirguidelinesonwhatconstitutesapositivefinding,disregardingsignificanttrendtestswhennocorrespondingpairwisecomparisonsarealsosignificant;

• Disregardingpositivefindingsindosesthatareclearlynotabovethemaximumdosetheanimalscouldbegivenwithcompromisingtheintegrityofthestudy;

• Usingunreasonableargumentsabouttheoverallfalsepositiveratesinthestudywithoutactuallydoingananalysisofthisissue;

• Failingtorecognizethesimilarfindingsinsimilarstudiesandtodoapooledanalysistodetermineifthenegativeeffectsinonestudycanceloutthepositiveeffectsinanother;

• Givingverylittleweighttostudiesfromtheliteratureandrelyingalmostentirelyonstudiesprovidedbyindustrythathavenotundergonepeerreviewforbothqualityand,moreimportantlyinsomecases,interpretationofthefindings;and

• Comparingresultsacrossdifferentspeciesandstrainsfortheanimalcancerstudiesandthemechanisticstudieswithlittleregardforuniquefindingsinanyonestudyandconsistentfindingsacrossmultiplestudies.

SimilarcommentsapplytotheevaluationdonebytheEuropeanFoodSafetyAuthority[89]andtheEuropeanChemicalAgency[233].Mydetailedcommentstothese


81

agenciesontheirriskassessmentsareattached.TherewerecommentstomycommentstoEPAbyotherscientistsandIalsorespondedtothosecommentsintheEPAdocketforglyphosate.ThesearealsoincludedintheattachedAppendices.

________________________________________ Dr.ChristopherJ.PortierCitedReferences

1. Portier,C.andD.Hoel,Optimaldesignofthechronicanimalbioassay.JToxicol

EnvironHealth,1983.12(1):p.1-19.2. Portier,C.J.andD.G.Hoel,DesignoftheChronicAnimalBioassayforGoodness

ofFittoMultistageModels.Biometrics,1983.39(3):p.809-809.3. Bailer,A.J.andC.J.Portier,Effectsoftreatment-inducedmortalityandtumor-

inducedmortalityontestsforcarcinogenicityinsmallsamples.Biometrics,1988.44(2):p.417-31.

4. Portier,C.J.andA.J.Bailer,Testingforincreasedcarcinogenicityusingasurvival-adjustedquantalresponsetest.FundamApplToxicol,1989.12(4):p.731-7.

5. Portier,C.J.,J.C.Hedges,andD.G.Hoel,Age-specificmodelsofmortalityandtumoronsetforhistoricalcontrolanimalsintheNationalToxicologyProgram'scarcinogenicityexperiments.CancerRes,1986.46(9):p.4372-8.

6. Portier,C.J.andA.J.Bailer,2-StageModelsofTumor-IncidenceforHistoricalControlAnimalsintheNationalToxicologyProgramsCarcinogenicityExperiments.JournalofToxicologyandEnvironmentalHealth,1989.27(1):p.21-45.

7. Portier,C.J.andL.Edler,Two-stagemodelsofcarcinogenesis,classificationofagents,anddesignofexperiments.FundamApplToxicol,1990.14(3):p.444-60.

8. Portier,C.J.,etal.,Biologicallybasedmodelsforriskassessment.IARCSciPubl,1990(104):p.20-8.

9. Kopp-Schneider,A.andC.J.Portier,Distinguishingbetweenmodelsofcarcinogenesis:theroleofclonalexpansion.FundamApplToxicol,1991.17(3):p.601-13.

10. Kopp-Schneider,A.,C.J.Portier,andF.Rippmann,TheapplicationofamultistagemodelthatincorporatesDNAdamageandrepairtotheanalysisofinitiation/promotionexperiments.MathBiosci,1991.105(2):p.139-66.

11. Kopp-Schneider,A.,C.J.Portier,andF.Rippmann,TheApplicationofaMultistageModelThatIncorporatesDNADamageandRepairtotheAnalysisofInitiationPromotionExperiments.MathematicalBiosciences,1991.105(2):p.139-166.

12. Kopp-Schneider,A.andC.J.Portier,BirthandDeathDifferentiationRatesofPapillomasinMouseSkin.Carcinogenesis,1992.13(6):p.973-978.


82

13. Portier,C.J.,A.Kopp-Schneider,andC.D.Sherman,UsingCellReplicationDatainMathematical-ModelinginCarcinogenesis.EnvironmentalHealthPerspectives,1993.101:p.79-86.

14. Kopp-Schneider,A.,C.J.Portier,andC.D.Sherman,TheExactFormulaforTumor-Incidenceinthe2-StageModel.RiskAnalysis,1994.14(6):p.1079-1080.

15. Portier,C.,etal.,Modelingthenumberandsizeofhepaticfocallesionsfollowingexposureto2378-TCDD.OrganohalogenCompounds,1994.21:p.393-397.

16. Kopp-Schneider,A.andC.J.Portier,CarcinomaformationinNMRImouseskinpaintingstudiesisaprocesssuggestinggreaterthantwostages.Carcinogenesis,1995.16(1):p.53-9.

17. Portier,C.J.,A.Kopp-Schneider,andC.D.Sherman,Calculatingtumorincidenceratesinstochasticmodelsofcarcinogenesis.MathematicalBiosciences,1996.135(2):p.129-146.

18. Portier,C.J.,etal.,Modelingthenumberandsizeofhepaticfocallesionsfollowingexposureto2,3,7,8-TCDD.ToxicologyandAppliedPharmacology,1996.138(1):p.20-30.

19. Kopp-Schneider,A.,C.Portier,andP.Bannasch,Amodelforhepatocarcinogenesistreatingphenotypicalchangesinfocalhepatocellularlesionsasepigeneticevents.MathBiosci,1998.148(2):p.181-204.

20. Toyoshiba,H.,etal.,Geneinteractionnetworkanalysissuggestsdifferencesbetweenhighandlowdosesofacetaminophen.ToxicolApplPharmacol,2006.215(3):p.306-16.

21. Gohlke,J.M.andC.J.Portier,Theforestforthetrees:asystemsapproachtohumanhealthresearch.EnvironHealthPerspect,2007.115(9):p.1261-3.

22. Gohlke,J.M.,etal.,Geneticandenvironmentalpathwaystocomplexdiseases.BMCSystBiol,2009.3:p.46.

23. Gohlke,J.M.,etal.,Estimatingtheglobalpublichealthimplicationsofelectricityandcoalconsumption.EnvironHealthPerspect,2011.119(6):p.821-6.

24. Aylward,L.L.,etal.,EvaluationofbiomonitoringdatafromtheCDCNationalExposureReportinariskassessmentcontext:perspectivesacrosschemicals.EnvironHealthPerspect,2013.121(3):p.287-94.

25. Trong,l.andC.Portier.ProceedingsoftheVietNam–UnitedStatesScientificConferencesonHumanHealthandEnvironmentalEffectsofAgentOrange/Dioxin,Part1and2.2002.HaNoi,Vietnam:USNationalInstituteofEnvironmentalSciencesandtheGovernmentofVietnam.

26. Portier,C.J.andM.S.Wolfe,eds.EMFScienceReviewSymposiumBreakoutGroupReportforEpidemiologyResearchFindings.1998,NationalInstituteofEnvironmentalHealthSciences:ResearchTrianglePark,NorthCarolina.

27. Portier,C.J.andM.S.Wolfe,eds.EMFScienceReviewSymposiumBreakoutGroupReportforClinicalandInVivoLaboratoryFindings.NIHPublicationNumber98-4400.1998,NationalInstituteofEnvironmentalHealthSciences:ResearchTrianglePark,NorthCarolina.

28. Portier,C.J.andM.S.Wolfe,eds.AssessmentofHealthEffectsfromExposuretoPower-LineFrequencyElectricandMagneticFields.NIHPublicationNumber98-


83

3981.1998,NationalInstituteofEnvironmentalHealthSciences:ResearchTrianglePark,NorthCarolina.508.

29. Portier,C.J.,etal.,AHumanHealthPerspectiveonClimateChange,N.I.o.E.H.S.HealthandHumanServices,Editor.2010,EnvironmentalHealthPerspectives:ResearchTrianglePark,NC72.

30. IARC.PreambletotheIARCMonographs.2006[cited20173/8/17];Availablefrom:http://monographs.iarc.fr/ENG/Preamble/CurrentPreamble.pdf.

31. NationalResearchCouncil.ScienceandJudgmentinRiskAssessment.1994,Washington(DC):NationalAcademyofSciencesPress.

32. NationalResearchCouncil.ScienceandDecisions:AdvancingRiskAssessment.2009,Washington(DC):NationalAcademiesPress.

33. USEPA,GuidelinesforCarcinogenRiskAssessment,U.E.P.Agency,Editor.2005:WashingtonDC166.

34. EuropeanChemicalsAgency,GuidanceontheApplicationoftheCLPCriteria:GuidancetoRegulation(EC)No1272/2008onclassification,labellingandpackaging(CLP)ofsubstancesandmixtures.2015,EuropeanChemicalsAgency:Helsinki,Finland.

35. NationalToxicologyProgram.HandbookforPreparingReportonCarcinogensMonographs.2015[cited20173/8/17];Availablefrom:https://ntp.niehs.nih.gov/ntp/roc/handbook/roc_handbook_508.pdf.

36. Hill,A.B.,TheEnvironmentandDisease:AssociationorCausation?ProcRSocMed,1965.58:p.295-300.

37. Smith,M.T.,etal.,KeyCharacteristicsofCarcinogensasaBasisforOrganizingDataonMechanismsofCarcinogenesis.EnvironHealthPerspect,2016.124(6):p.713-21.

38. Chang,E.T.andE.Delzell,Systematicreviewandmeta-analysisofglyphosateexposureandriskoflymphohematopoieticcancers.JEnvironSciHealthB,2016:p.1-27.

39. Cantor,K.P.,etal.,Pesticidesandotheragriculturalriskfactorsfornon-Hodgkin'slymphomaamongmeninIowaandMinnesota.CancerRes,1992.52(9):p.2447-2455.

40. Nordstrom,M.,etal.,Occupationalexposures,animalexposureandsmokingasriskfactorsforhairycellleukaemiaevaluatedinacase-controlstudy.Br.JCancer,1998.77(11):p.2048-2052.

41. Hardell,L.andM.Eriksson,Acase-controlstudyofnon-Hodgkinlymphomaandexposuretopesticides.Cancer,1999.85(6):p.1353-1360.

42. Hardell,L.,M.Eriksson,andM.Nordstrom,Exposuretopesticidesasriskfactorfornon-Hodgkin'slymphomaandhairycellleukemia:pooledanalysisoftwoSwedishcase-controlstudies.Leuk.Lymphoma,2002.43(5):p.1043-1049.

43. DeRoos,A.J.,etal.,Integrativeassessmentofmultiplepesticidesasriskfactorsfornon-Hodgkin'slymphomaamongmen.OccupEnvironMed,2003.60(9):p.E11.

44. Lee,W.J.,etal.,Non-Hodgkin'slymphomaamongasthmaticsexposedtopesticides.IntJCancer,2004.111(2):p.298-302.


84

45. DeRoos,A.J.,etal.,Cancerincidenceamongglyphosate-exposedpesticideapplicatorsintheAgriculturalHealthStudy.EnvironHealthPerspect,2005.113(1):p.49-54.

46. Eriksson,M.,etal.,Pesticideexposureasriskfactorfornon-Hodgkinlymphomaincludinghistopathologicalsubgroupanalysis.IntJCancer,2008.123(7):p.1657-63.

47. Orsi,L.,etal.,Occupationalexposuretopesticidesandlymphoidneoplasmsamongmen:resultsofaFrenchcase-controlstudy.OccupEnvironMed,2009.66(5):p.291-8.

48. Hohenadel,K.,etal.,Exposuretomultiplepesticidesandriskofnon-HodgkinlymphomainmenfromsixCanadianprovinces.IntJEnvironResPublicHealth,2011.8(6):p.2320-30.

49. Cocco,P.,etal.,Lymphomariskandoccupationalexposuretopesticides:resultsoftheEpilymphstudy.OccupEnvironMed,2013.70(2):p.91-8.

50. McDuffie,H.H.,etal.,Non-Hodgkin'slymphomaandspecificpesticideexposuresinmen:cross-Canadastudyofpesticidesandhealth.CancerEpidemiolBiomarkersPrev,2001.10(11):p.1155-1163.

51. Zahm,S.H.,etal.,Acase-controlstudyofnon-Hodgkin'slymphomaandtheherbicide2,4-dichlorophenoxyaceticacid(2,4-D)ineasternNebraska.Epidemiology,1990.1(5):p.349-356.

52. Hoar,S.K.,etal.,Agriculturalherbicideuseandriskoflymphomaandsoft-tissuesarcoma.JAMA,1986.256(9):p.1141-1147.

53. Blair,A.,etal.,Reliabilityofreportingonlife-styleandagriculturalfactorsbyasampleofparticipantsintheAgriculturalHealthStudyfromIowa.Epidemiology,2002.13(1):p.94-9.

54. FIFRAScientificAdvisoryPanelMeetingMinutes,MeetingMinutesandFinalReportoftheDecember13-16,2016FIFRASAPMeetingHeldtoConsiderandReviewScientificIssuesAssociatedwithEPA'sEvaluationoftheCarcinogenicPotentialofGlyphasate,O.o.P.Programs,Editor.2017,USEnvironmentalProtectionAgency:Washington,DC101.

55. Portier,C.J.,etal.,DifferencesinthecarcinogenicevaluationofglyphosatebetweentheInternationalAgencyforResearchonCancer(IARC)andtheEuropeanFoodSafetyAuthority(EFSA).JEpidemiolCommunityHealth,2016.70(8):p.741-5.

56. IARCWorkingGroup,Glyphosate,In:SomeOrganophosphateInsecticidesandHerbicides:Diazinon,Glyphosate,Malathion,Parathion,andTetrachlorvinphos.2015,IARCMonogrProg.V.112.p.1-92.

57. Schinasi,L.andM.E.Leon,Non-Hodgkinlymphomaandoccupationalexposuretoagriculturalpesticidechemicalgroupsandactiveingredients:asystematicreviewandmeta-analysis.IntJEnvironResPublicHealth,2014.11(4):p.4449-527.

58. Ioannidis,J.P.,Interpretationoftestsofheterogeneityandbiasinmeta-analysis.JEvalClinPract,2008.14(5):p.951-7.

59. OECD,GuidanceDocument116ontheConductandDesignofChronicToxicityandCarcinogenicityStudies,H.a.S.P.Environment,Editor.2012,OECD:Paris.


85

60. Haseman,J.K.,Statisticalissuesinthedesign,analysisandinterpretationofanimalcarcinogenicitystudies.EnvironHealthPerspect,1984.58:p.385-92.

61. EnvironmentalProtectionAgency,GlyphosateIssuePaper:EvaluationofCarcinogenicPotential.2016:Washington,DC.

62. NRCCommitteetoReviewtheStyreneAssessmentintheNationalToxicologyProgram12thReportonCarcinogens,inReviewoftheStyreneAssessmentintheNationalToxicologyProgram12thReportonCarcinogens:WorkshopSummary.2014,NationalAcademiesPress:Washington(DC).

63. Bieler,G.S.andR.L.Williams,Ratioestimates,thedeltamethod,andquantalresponsetestsforincreasedcarcinogenicity.Biometrics,1993.49(3):p.793-801.

64. Piegorsch,W.W.andA.J.Bailer,Analyzingenvironmentaldata.2005,Chichester,WestSussex,England;Hoboken,NJ:Wiley.xv,496p.

65. Haseman,J.K.,G.A.Boorman,andJ.Huff,Valueofhistoricalcontroldataandotherissuesrelatedtotheevaluationoflong-termrodentcarcinogenicitystudies.ToxicolPathol,1997.25(5):p.524-7.

66. Haseman,J.K.,J.Huff,andG.A.Boorman,Useofhistoricalcontroldataincarcinogenicitystudiesinrodents.ToxicolPathol,1984.12(2):p.126-35.

67. OECD,CombinedChronicToxicity\CarcinogenicityStudies,OECDGuidelinesforTestingofChemicals,No.453,OECD,Editor.2009:Paris.

68. Atkinson,C.,Strutt,A.,Henderson,W.,etal.,104-WeekChronicFeeding/Oncogenicitystudyinratswith52-weekinterimkill.1993.

69. Brammer.,GlyphosateAcid:TwoYearDietaryToxicityandOncogenicityStudyinWistarRats.2001:CentralToxicologyLaboratory,AlderleyParkMacclesfield,Cheshire,UK.

70. Burnett,P.,Borders,J.;,Kush,J.,ReporttoMonsantoCompany:TwoYearChronicOralToxicityStudywithCP-76100inAlbinoRats.1979,IndustrialBio-TestLaboratories,Inc.:onbehalfofMonsantoCo.

71. Chruscielska,K.B.,J.;Kita,K,etal.,Glyphosate:Evaluationofchronicactivityandpossiblefar-reachingeffects-Part1.Studiesonchronictoxicity.Pestycydy,2000.3-4:p.10.

72. Enemoto,K.,24-MonthOralChronicToxicityandOncogenicityStudyinRats,Vol.1.1997:TheInstituteofEnvironmentalToxicology,Kodaira-shi,Tokyo,Japan.

73. Excel,Combinedchronictoxicity/carcinogenicitystudyofglyphosatetechnicalinSpragueDawleyrats..1997:IndianInstituteofToxicology,Pune,India.

74. Lankas,G.,P.,ALifetimeStudyofGlyphosateinRats.1981,Monsanto:ReportNo.77-2062preparedbyBioDynamics,Inc.

75. Pavkov,K.L.,Wyand,S.,Two-YearChronicToxicityandOncogenecityDietaryStudywithSC-0224inRats.1987:StaufferChemicalCompany.

76. Reyna,M.S.,Gordon,D.E.,Two-YearChronicOralToxicityStudywithCP67573inAlbinoRats:Monsanto.1974.

77. Seralini,G.E.,etal.,Republishedstudy:long-termtoxicityofaRoundupherbicideandaRoundup-tolerantgeneticallymodifiedmaize.EnvironmentalSciencesEurope,2014.26(1):p.14.


86

78. Stout,L.D.a.R.,P.A.,ChronicStudyofGlyphosateAdministeredinFeedtoAlbinoRats.1990:Monsanto.

79. Suresh,T.P.,CombinedChronicToxicityandCarcinogenicityStudywithGlyphosateTechnicalinWistarRats.1996,Syngenta:ToxicologyDepartmentRallisResearchCentre,RallisIndiaLimited.

80. Wood,E.,Dunster,J.,Watson,P.,andBrooks,P.,GlyphosateTechnical:DietaryCombinedChronicToxicity/CarcinogenicityStudyintheRat.2009:HarlanLaboratoriesLimited,ShardlowBusinessPark,Shardlow,DerbyshireDE722GD,UK.StudyNo.2060-012.April,23,2009.

81. Atkinson,C.,Martin,T.,Hudson,P.,andRobb,D.,Glyphosate:104weekdietarycarcinogenicitystudyinmice.1993:InvereskResearchInternational,Tranent,EH332NE,Scotland.IRIProjectNo.438618.April7,1993.

82. George,J.,etal.,Studiesonglyphosate-inducedcarcinogenicityinmouseskin:aproteomicapproach.JProteomics,2010.73(5):p.951-64.

83. Knezevich,A.L.andG.K.Hogan,Achronicfeedingstudyofglyphosateinmice:Monsanto.1983:Bio/DynamicInc.,datedJuly21,1983.ReportNo.77-2011..

84. Kumar,D.P.S.,CarcinogenicityStudywithGlyphosateTechnicalinSwissAlbinoMice.2001:ToxicologyDepartmentRallisResearchCentre,RallisIndiaLimited.StudyNo.TOXI:1559.CARCI-M.

85. Pavkov,K.L.,Turnier,J.C.,Two-YearChronicToxicityandOncogenecityDietaryStudywithSC-0224inMice.1987:StaufferChemicalCompany.

86. Reyna,M.S.andD.E.Gordon,18-MonthCarcinogenicityStudywithCP67573inSwissWhiteMice:Monsanto.1973:IndustrialBio-TestLaboratories,Inc.

87. Sugimoto,K.,18-MonthOralOncogenicityStudyinMice,Vol.1and2.1997:TheInstituteofEnvironmentalToxicology,2-772,Suzuki-cho,Kodaira-shi,Tokyo,187,Japan,StudyNo.:IET94-0151.

88. Wood,E.,Dunster,J.,Watson,P.,andBrooks,P.,GlyphosateTechnical:DietaryCarcinogenicityStudyintheMouse.2009:HarlanLaboratoriesLimited,ShardlowBusinessPark,Shardlow,DerbyshireDE722GD,UK.StudyNo.2060-011.April,22,2009.

89. (BfR).,G.F.I.f.R.A.,GlyphosateRenewalAssessmentReport:ToxicologyandMetabolism.2013.V.3.p.

90. (BfR).,G.F.I.f.R.A.,GlyphosateRenewalAssessmentReport:AssessmentofIARCMonographiesVolume112(2015).2015.V.3GlyphosateAddendumItoRAR.p.

91. Greim,H.,etal.,Evaluationofcarcinogenicpotentialoftheherbicideglyphosate,drawingontumorincidencedatafromfourteenchronic/carcinogenicityrodentstudies.CritRevToxicol,2015.45(3):p.185-208.

92. Blettner,M.,etal.,Traditionalreviews,meta-analysesandpooledanalysesinepidemiology.IntJEpidemiol,1999.28(1):p.1-9.

93. Friedenreich,C.M.,Methodsforpooledanalysesofepidemiologicstudies.Epidemiology,1993.4(4):p.295-302.

94. Portier,C.J.,CommentsonGlyphosateIssuePaper(EPA-HQ-OPP-2016-0385-0094).2016,USEPADocketNumber(EPA-HQ-OPP-2016-0385-0094):WashingtonDC.


87

95. Hogan,g.k.,AchronicFeedingStudyofGlyphosate(RoundupTechnical)inRats.1981,Monsanto:Bio/dynamicsInc.,E.Millstone,NJ.

96. EPA,EPAMemoStoutandRuecker,I.WilliamDykstra.ToxicologyBranch,Editor.1991.V.MRID416438-01Toxreview008897.p.

97. Uno,H.,etal.,AversatiletestforequalityoftwosurvivalfunctionsbasedonweighteddifferencesofKaplan-Meiercurves.StatMed,2015.34(28):p.3680-95.

98. Giknis,M.andC.Clifford.NeoplasticandNon-NeoplasticLesionsintheCharlesRiverWistarHannover[Crl:WI(Han)]Rat.2011;Availablefrom:http://www.criver.com/files/pdfs/rms/wistarhan/rm_rm_r_wistar_han_tox_data_2011.aspx.

99. Acquavella,J.,etal.,Glyphosateepidemiologyexpertpanelreview:aweightofevidencesystematicreviewoftherelationshipbetweenglyphosateexposureandnon-Hodgkin'slymphomaormultiplemyeloma.CritRevToxicol,2016.46(sup1):p.28-43.

100. Lacayo,H.,Memorandum:UseofhistoricaldataindeterminingtheweightofevidencefromkidneytumorincidenceintheGlyphosatetwo-yearfeedingstudyandsomeremarksonfalsepositives,S.M.S.Branch,Editor.1985,USEPA:Washington,DCC.

101. Chandra,M.andC.H.Frith,SpontaneousneoplasmsinagedCD-1mice.ToxicolLett,1992.61(1):p.67-74.

102. Giknis,M.andC.Clifford,SpontaneousNeoplasticLesionsintheCrI:CD-1(ICR)BRMouse.2000,CharlesRiverLaboratories.

103. Giknis,M.andC.Clifford,SpontaneousNeoplasticLesionsintheCrI:CD-1(ICR)BRMouseinControlGroupsfrom18Monthto2yearStudies.2005,CharlesRiverLaboratories.

104. Eustis,S.L.,etal.,Theutilityofmultiple-sectionsamplinginthehistopathologicalevaluationofthekidneyforcarcinogenicitystudies.ToxicolPathol,1994.22(5):p.457-72.

105. Kasza,L.,Memorandum:Glyphosate-EvaluationofKidneyTumorsinMaleMice.ChronicFeedingStudy,T.B.Branch,Editor.1985,USEPA:Washington,DCC.

106. deVivarChevez,A.R.,J.Finke,andR.Bukowski,Theroleofinflammationinkidneycancer.AdvExpMedBiol,2014.816:p.197-234.

107. Dixon,D.,etal.,SummaryofchemicallyinducedpulmonarylesionsintheNationalToxicologyProgram(NTP)toxicologyandcarcinogenesisstudies.ToxicolPathol,2008.36(3):p.428-39.

108. Son,W.C.andC.Gopinath,EarlyoccurrenceofspontaneoustumorsinCD-1miceandSprague-Dawleyrats.ToxicolPathol,2004.32(4):p.371-4.

109. OECD,CarcinogenicityStudies,OECDGuidelinefortheTestingofChemicals,No.451,OECD,Editor.2009:Paris.

110. Martens,M.,MemorandumtoJoeHaseman,March2,2017:AnalysistumordataRAC40meeting,G.T.Force,Editor.2017:Helsinki,Finland.

111. Haseman,J.K.,CommentsonthelaboratoryanimalcarcinogenicitystudiesevaluatedinEPA’sGlyphosateIssuePaper:EvaluationofCarcinogenicPotential,


88

andanassessmentofthecommentsbyDr.ChristopherPortieronthesestudies.2016,USEPADocketNumber(EPA-HQ-OPP-2016-0385-0094):WashingtonDC.

112. Hanahan,D.andR.A.Weinberg,Hallmarksofcancer:thenextgeneration.Cell,2011.144(5):p.646-74.

113. Higgins,J.P.T.andS.Green,CochraneHandbookforSystematicReviewsofInterventions.2011,TheCochraneCollaboration.

114. Birnbaum,L.S.,etal.,ImplementingsystematicreviewattheNationalToxicologyProgram:statusandnextsteps.EnvironHealthPerspect,2013.121(4):p.A108-9.

115. Murray,H.E.andK.A.Thayer,Implementingsystematicreviewintoxicologicalprofiles:ATSDRandNIEHS/NTPcollaboration.JEnvironHealth,2014.76(8):p.34-5.

116. Rooney,A.A.,etal.,Systematicreviewandevidenceintegrationforliterature-basedenvironmentalhealthscienceassessments.EnvironHealthPerspect,2014.122(7):p.711-8.

117. Vandenberg,L.N.,etal.,Aproposedframeworkforthesystematicreviewandintegratedassessment(SYRINA)ofendocrinedisruptingchemicals.EnvironHealth,2016.15(1):p.74.

118. Paz-y-Miño,C.,etal.,EvaluationofDNAdamageinanEcuadorianpopulationexposedtoglyphosate.GeneticsandMolecularBiology,2007.30(2).

119. Paz-y-Miño,C.,etal.,Baselinedeterminationinsocial,health,andgeneticareasincommunitiesaffectedbyglyphosateaerialsprayingonthenortheasternEcuadorianborder.ReviewsonEnvironmentalHealth,2011.26(1).

120. Bolognesi,C.,etal.,Biomonitoringofgenotoxicriskinagriculturalworkersfromfivecolombianregions:associationtooccupationalexposuretoglyphosate.JToxicolEnvironHealthA,2009.72(15-16):p.986-97.

121. Kier,L.D.,Reviewofgenotoxicitybiomonitoringstudiesofglyphosate-basedformulations.CritRevToxicol,2015.45(3):p.209-18.

122. Mladinic,M.,etal.,Evaluationofgenomedamageanditsrelationtooxidativestressinducedbyglyphosateinhumanlymphocytesinvitro.EnvironMolMutagen,2009.50(9):p.800-7.

123. Alvarez-Moya,C.,etal.,Comparisonoftheinvivoandinvitrogenotoxicityofglyphosateisopropylaminesaltinthreedifferentorganisms.Genet.MolBiol,2014.37(1):p.105-110.

124. Manas,F.,etal.,Genotoxicityofglyphosateassessedbythecometassayandcytogenetictests.EnvironToxicolPharmacol,2009.28(1):p.37-41.

125. Monroy,C.M.,etal.,Citotoxicidadygenotoxicidadencélulashumanasexpuestasinvitroaglifosato.Biomédica,2005.25:p.335-45.

126. Lueken,A.,etal.,SynergisticDNAdamagebyoxidativestress(inducedbyH2O2)andnongenotoxicenvironmentalchemicalsinhumanfibroblasts.ToxicologyLetters,2004.147(1):p.35-43.

127. Koller,V.J.,etal.,CytotoxicandDNA-damagingpropertiesofglyphosateandRoundupinhuman-derivedbuccalepithelialcells.ArchToxicol,2012.86(5):p.805-13.


89

128. Gasnier,C.,etal.,Glyphosate-basedherbicidesaretoxicandendocrinedisruptorsinhumancelllines.Toxicology,2009.262(3):p.184-91.

129. Mladinic,M.,P.Perkovic,andD.Zeljezic,Characterizationofchromatininstabilitiesinducedbyglyphosate,terbuthylazineandcarbofuranusingcytomeFISHassay.ToxicolLett,2009.189(2):p.130-7.

130. Bolognesi,C.,etal.,GenotoxicActivityofGlyphosateandItsTechnicalFormulationRoundup.J.Agric.FoodChem.,1997.45:p.1957-1962.

131. Lioi,M.B.,etal.,Genotoxicityandoxidativestressinducedbypesticideexposureinbovinelymphocyteculturesinvitro.MutatRes,1998.403(1-2):p.13-20.

132. Vigfusson,N.V.andE.R.Vyse,Theeffectofthepesticides,Dexon,CaptanandRoundup,onsister-chromatidexchangesinhumanlymphocytesinvitro.MutatRes,1980.79(1):p.53-57.

133. Peluso,M.,etal.,32P-postlabelingdetectionofDNAadductsinmicetreatedwiththeherbicideRoundup.EnvironMolMutagen,1998.31(1):p.55-59.

134. Rank,J.,etal.,GenotoxicitytestingoftheherbicideRoundupanditsactiveingredientglyphosateisopropylamineusingthemousebonemarrowmicronucleustest,Salmonellamutagenicitytest,andAlliumanaphase-telophasetest.MutatRes,1993.300(1):p.29-36.

135. Dimitrov,B.D.,etal.,ComparativegenotoxicityoftheherbicidesRoundup,StompandRegloneinplantandmammaliantestsystems.Mutagenesis,2006.21(6):p.375-82.

136. Prasad,S.,etal.,Clastogeniceffectsofglyphosateinbonemarrowcellsofswissalbinomice.JToxicol,2009.2009:p.308985.

137. Grisolia,C.K.,Acomparisonbetweenmouseandfishmicronucleustestusingcyclophosphamide,mitomycinCandvariouspesticides.MutatRes,2002.518(2):p.145-150.

138. CoutinhodoNascimentoAandG.C.,ComparativeanalysisbetweenmicronucleitestsinmiceandinperipheralerythrocytesofOreochromisniloticusinevaluationofmutagenicpotentialoftheagrotoxinsdeltamethrin,dicofol,glyphosate,andimazapyr.Pesticides:REcotoxicolEMeioAmbiente,Curitiba,2000.10:p.8.

139. Cavusoglu,K.,etal.,ProtectiveeffectofGinkgobilobaL.leafextractagainstglyphosatetoxicityinSwissalbinomice.JMedFood,2011.14(10):p.1263-72.

140. Chan,P.andJ.Mahler,ToxicityStudiesofGlyphosate(CASNo.1071836)AdministeredinDosedFeedtoF344/NRatsandB6C3F1Mice,N.T.Program,Editor.1992:ResearchTrianglePark,NC.

141. Li,A.P.andT.J.Long,Anevaluationofthegenotoxicpotentialofglyphosate.Fundam.ApplToxicol,1988.10(3):p.537-546.

142. Roustan,A.,etal.,Genotoxicityofmixturesofglyphosateandatrazineandtheirenvironmentaltransformationproductsbeforeandafterphotoactivation.Chemosphere,2014.108:p.93-100.

143. Sivikova,K.andJ.Dianovsky,Cytogeneticeffectoftechnicalglyphosateoncultivatedbovineperipherallymphocytes.IntJHygEnvironHealth,2006.209(1):p.15-20.


90

144. HOLEČKOVÁ,B.,EVALUATIONOFTHEINVITROEFFECTOFGLYPHOSATE-BASEDHERBICIDEONBOVINELYMPHOCYTESUSINGCHROMOSOMEPAINTING.BullVetInstPulawy,2006.50:p.533-536.

145. Moreno,N.C.,S.H.Sofia,andC.B.Martinez,GenotoxiceffectsoftheherbicideRoundupTransorbanditsactiveingredientglyphosateonthefishProchiloduslineatus.EnvironToxicolPharmacol,2014.37(1):p.448-54.

146. Lopes,F.M.,etal.,EffectofglyphosateonthespermqualityofzebrafishDaniorerio.AquatToxicol,2014.155:p.322-6.

147. Guilherme,S.,etal.,AreDNA-damagingeffectsinducedbyherbicideformulations(Roundup(R)andGarlon(R))infishtransientandreversibleuponcessationofexposure?AquatToxicol,2014.155:p.213-21.

148. Akcha,F.,C.Spagnol,andJ.Rouxel,Genotoxicityofdiuronandglyphosateinoysterspermatozoaandembryos.AquatToxicol,2012.106-107:p.104-13.

149. Kaya,B.,etal.,UseoftheDrosophilawingspottestinthegenotoxicitytestingofdifferentherbicides.EnvironMolMutagen,2000.36(1):p.40-6.

150. Cavalcante,D.G.,C.B.Martinez,andS.H.Sofia,GenotoxiceffectsofRounduponthefishProchiloduslineatus.MutatRes,2008.655(1-2):p.41-6.

151. Cavas,T.andS.Konen,DetectionofcytogeneticandDNAdamageinperipheralerythrocytesofgoldfish(Carassiusauratus)exposedtoaglyphosateformulationusingthemicronucleustestandthecometassay.Mutagenesis,2007.22(4):p.263-8.

152. deCastilhosGhisi,N.andM.M.Cestari,GenotoxiceffectsoftheherbicideRoundup((R))inthefishCorydoraspaleatus(Jenyns1842)aftershort-term,environmentallylowconcentrationexposure.EnvironMonitAssess,2013.185(4):p.3201-7.

153. DeSouzaFilho,J.,etal.,MutagenicityandgenotoxicityingillerythrocytecellsofPoeciliareticulataexposedtoaglyphosateformulation.BullEnvironContamToxicol,2013.91(5):p.583-7.

154. Gholami-Seyedkolaei,S.J.,etal.,Optimizationofrecoverypatternsincommoncarpexposedtoroundupusingresponsesurfacemethodology:evaluationofneurotoxicityandgenotoxicityeffectsandbiochemicalparameters.EcotoxicolEnvironSaf,2013.98:p.152-61.

155. Guilherme,S.,etal.,Europeaneel(Anguillaanguilla)genotoxicandpro-oxidantresponsesfollowingshort-termexposuretoRoundup--aglyphosate-basedherbicide.Mutagenesis,2010.25(5):p.523-30.

156. Guilherme,S.,etal.,DifferentialgenotoxicityofRoundup((R))formulationanditsconstituentsinbloodcellsoffish(Anguillaanguilla):considerationsonchemicalinteractionsandDNAdamagingmechanisms.Ecotoxicology,2012.21(5):p.1381-90.

157. Marques,A.,etal.,ProgressionofDNAdamageinducedbyaglyphosate-basedherbicideinfish(Anguillaanguilla)uponexposureandpost-exposureperiods--insightsintothemechanismsofgenotoxicityandDNArepair.CompBiochemPhysiolCToxicolPharmacol,2014.166:p.126-33.


91

158. Marques,A.,etal.,ErratumTo:ProgressionofDNAdamageinducedbyaglyphosate-basedherbicideinfish(Anguillaanguilla)uponexposureandpost-exposureperiods--insightsintothemechanismsofgenotoxicityandDNArepair.CompBiochemPhysiolCToxicolPharmacol,2015.168:p.1.

159. Nwani,C.D.,etal.,DNAdamageandoxidativestressmodulatoryeffectsofglyphosate-basedherbicideinfreshwaterfish,Channapunctatus.EnvironToxicolPharmacol,2013.36(2):p.539-47.

160. Vera-Candioti,J.,S.Soloneski,andM.L.Larramendy,Evaluationofthegenotoxicandcytotoxiceffectsofglyphosate-basedherbicidesinthetenspottedlive-bearerfishCnesterodondecemmaculatus(Jenyns,1842).EcotoxicolEnvironSaf,2013.89:p.166-73.

161. Poletta,G.L.,etal.,Genetic,enzymaticanddevelopmentalalterationsobservedinCaimanlatirostrisexposedinovotopesticideformulationsandmixturesinanexperimentsimulatingenvironmentalexposure.EcotoxicolEnvironSaf,2011.74(4):p.852-9.

162. Poletta,G.L.,etal.,GenotoxicityoftheherbicideformulationRoundup(glyphosate)inbroad-snoutedcaiman(Caimanlatirostris)evidencedbytheCometassayandtheMicronucleustest.MutatRes,2009.672(2):p.95-102.

163. Clements,C.,S.Ralph,andM.Petras,GenotoxicityofselectherbicidesinRanacatesbeianatadpolesusingthealkalinesingle-cellgelDNAelectrophoresis(comet)assay.EnvironMolMutagen,1997.29(3):p.277-288.

164. Meza-Joya,F.L.,M.P.Ramirez-Pinilla,andJ.L.Fuentes-Lorenzo,Toxic,cytotoxic,andgenotoxiceffectsofaglyphosateformulation(Roundup(R)SL-Cosmoflux(R)411F)inthedirect-developingfrogEleutherodactylusjohnstonei.EnvironMolMutagen,2013.54(5):p.362-73.

165. Yadav,S.S.,etal.,ToxicandgenotoxiceffectsofRoundupontadpolesoftheIndianskitteringfrog(Euflictiscyanophlyctis)inthepresenceandabsenceofpredatorstress.AquatToxicol,2013.132-133:p.1-8.

166. dosSantos,K.C.andC.B.Martinez,GenotoxicandbiochemicaleffectsofatrazineandRoundup((R)),aloneandincombination,ontheAsianclamCorbiculafluminea.EcotoxicolEnvironSaf,2014.100:p.7-14.

167. Conners,D.E.andM.C.Black,EvaluationofLethalityandGenotoxicityintheFreshwaterMusselUtterbackiaimbecillis(Bivalvia:Unionidae)ExposedSinglyandinCombinationtoChemicalsUsedinLawnCare.ArchivesofEnvironmentalContaminationandToxicology,2004.46(3).

168. Mohamed,A.H.,SublethaltoxicityofRounduptoimmunologicalandmolecularaspectsofBiomphalariaalexandrinatoSchistosomamansoniinfection.EcotoxicolEnvironSaf,2011.74(4):p.754-60.

169. Muangphra,P.,W.Kwankua,andR.Gooneratne,Genotoxiceffectsofglyphosateorparaquatonearthwormcoelomocytes.EnvironToxicol,2014.29(6):p.612-20.

170. Piola,L.,etal.,Comparativetoxicityoftwoglyphosate-basedformulationstoEiseniaandreiunderlaboratoryconditions.Chemosphere,2013.91(4):p.545-51.

171. Kale,P.G.,etal.,Mutagenicitytestingofnineherbicidesandpesticidescurrentlyusedinagriculture.EnvironMolMutagen,1995.25(2):p.148-153.


92

172. Chen,L.,etal.,ThecombinedeffectsofUV-Bradiationandherbicidesonphotosynthesis,antioxidantenzymesandDNAdamageintwobloom-formingcyanobacteria.EcotoxicolEnvironSaf,2012.80:p.224-30.

173. Wang,G.,etal.,DamagetoDNAcausedbyUV-BradiationinthedesertcyanobacteriumScytonemajavanicumandtheeffectsofexogenouschemicalsontheprocess.Chemosphere,2012.88(4):p.413-7.

174. Alvarez-Moya,C.,etal.,EvaluationofgeneticdamageinducedbyglyphosateisopropylaminesaltusingTradescantiabioassays.Genet.MolBiol,2011.34(1):p.127-130.

175. Williams,G.M.,R.Kroes,andI.C.Munro,SafetyevaluationandriskassessmentoftheherbicideRoundupanditsactiveingredient,glyphosate,forhumans.RegulToxicolPharmacol,2000.31(2Pt1):p.117-65.

176. Moriya,M.,etal.,Furthermutagenicitystudiesonpesticidesinbacterialreversionassaysystems.MutatRes,1983.116(3-4):p.185-216.

177. Kier,L.D.andD.J.Kirkland,Reviewofgenotoxicitystudiesofglyphosateandglyphosate-basedformulations.CritRevToxicol,2013.43(4):p.283-315.

178. Lioi,M.B.,etal.,Cytogeneticdamageandinductionofpro-oxidantstateinhumanlymphocytesexposedinvitrotogliphosate,vinclozolin,atrazine,andDPX-E9636.EnvironMolMutagen,1998.32(1):p.39-46.

179. Chruscielska,K.B.,J.;Kita,K,etal.,Glyphosate:Evaluationofchronicactivityandpossiblefar-reachingeffects-Part2.StudiesonMutagenicActivity.Pestycydy,2000.3-4:p.5.

180. GhisiNde,C.,E.C.deOliveira,andA.J.Prioli,Doesexposuretoglyphosateleadtoanincreaseinthemicronucleifrequency?Asystematicandmeta-analyticreview.Chemosphere,2016.145:p.42-54.

181. Aghadavod,E.,etal.,RoleofOxidativeStressandInflammatoryFactorsinDiabeticKidneyDisease.IranJKidneyDis,2016.10(6):p.337-343.

182. Kamceva,G.,etal.,CigaretteSmokingandOxidativeStressinPatientswithCoronaryArteryDisease.OpenAccessMacedJMedSci,2016.4(4):p.636-640.

183. Qureshi,M.A.,Y.O.Kim,andD.Schuppan,Hepatocellularcarcinomainnonalcoholicfattyliverdisease:AlinkbetweenoxidativestressandT-cellsuppression.Hepatology,2016.64(5):p.1794-1797.

184. Sayanthooran,S.,etal.,UpregulationofOxidativeStressRelatedGenesinaChronicKidneyDiseaseAttributedtoSpecificGeographicalLocationsofSriLanka.BiomedResInt,2016.2016:p.7546265.

185. Turkmen,K.,Inflammation,oxidativestress,apoptosis,andautophagyindiabetesmellitusanddiabetickidneydisease:theFourHorsemenoftheApocalypse.IntUrolNephrol,2016.

186. Vakonaki,E.,etal.,ComplexinterplayofDNAdamage,DNArepairgenes,andoxidativestressincoronaryarterydisease.AnatolJCardiol,2016.16(12):p.939.

187. Hecht,F.,etal.,Theroleofoxidativestressonbreastcancerdevelopmentandtherapy.TumourBiol,2016.37(4):p.4281-91.

188. Li,L.andF.Chen,Oxidativestress,epigenetics,andcancerstemcellsinarseniccarcinogenesisandprevention.CurrPharmacolRep,2016.2(2):p.57-63.


93

189. Perse,M.,Oxidativestressinthepathogenesisofcolorectalcancer:causeorconsequence?BiomedResInt,2013.2013:p.725710.

190. Prasad,S.,etal.,OxidativeStressandCancer:AdvancesandChallenges.OxidMedCellLongev,2016.2016:p.5010423.

191. Toyokuni,S.,Oxidativestressasanicebergincarcinogenesisandcancerbiology.ArchBiochemBiophys,2016.595:p.46-9.

192. Kwiatkowska,M.,B.Huras,andB.Bukowska,Theeffectofmetabolitesandimpuritiesofglyphosateonhumanerythrocytes(invitro).PesticBiochemPhysiol,2014.109:p.34-43.

193. Chaufan,G.,I.Coalova,andC.RiosdeMolinaMdel,Glyphosatecommercialformulationcausescytotoxicity,oxidativeeffects,andapoptosisonhumancells:differenceswithitsactiveingredient.IntJToxicol,2014.33(1):p.29-38.

194. Coalova,I.,C.RiosdeMolinaMdel,andG.Chaufan,Influenceofthesprayadjuvantonthetoxicityeffectsofaglyphosateformulation.ToxicolInVitro,2014.28(7):p.1306-11.

195. Gehin,A.,etal.,VitaminsCandEreverseeffectofherbicide-inducedtoxicityonhumanepidermalcellsHaCaT:abiochemometricapproach.IntJPharm,2005.288(2):p.219-26.

196. Elie-Caille,C.,etal.,Morphologicaldamagesofaglyphosate-treatedhumankeratinocytecelllinerevealedbyamicro-tonanoscalemicroscopicinvestigation.CellBiolToxicol,2010.26(4):p.331-9.

197. Bonini,M.G.,etal.,Theoxidationof2',7'-dichlorofluorescintoreactiveoxygenspecies:aself-fulfillingprophesy?FreeRadicBiolMed,2006.40(6):p.968-75.

198. Kalyanaraman,B.,etal.,Measuringreactiveoxygenandnitrogenspecieswithfluorescentprobes:challengesandlimitations.FreeRadicBiolMed,2012.52(1):p.1-6.

199. George,J.andY.Shukla,EmptyingofIntracellularCalciumPoolandOxidativeStressImbalanceAreAssociatedwiththeGlyphosate-InducedProliferationinHumanSkinKeratinocytesHaCaTCells.ISRNDermatol,2013.2013:p.825180.

200. Valavanidis,A.,T.Vlachogianni,andC.Fiotakis,8-hydroxy-2'-deoxyguanosine(8-OHdG):Acriticalbiomarkerofoxidativestressandcarcinogenesis.JEnvironSciHealthCEnvironCarcinogEcotoxicolRev,2009.27(2):p.120-39.

201. Jasper,R.,etal.,Evaluationofbiochemical,hematologicalandoxidativeparametersinmiceexposedtotheherbicideglyphosate-Roundup((R)).InterdiscipToxicol,2012.5(3):p.133-40.

202. Astiz,M.,M.J.deAlaniz,andC.A.Marra,Antioxidantdefensesysteminratssimultaneouslyintoxicatedwithagrochemicals.EnvironToxicolPharmacol,2009.28(3):p.465-73.

203. Astiz,M.,M.J.deAlaniz,andC.A.Marra,Theoxidativedamageandinflammationcausedbypesticidesarerevertedbylipoicacidinratbrain.NeurochemInt,2012.61(7):p.1231-41.

204. Cattani,D.,etal.,Mechanismsunderlyingtheneurotoxicityinducedbyglyphosate-basedherbicideinimmaturerathippocampus:involvementofglutamateexcitotoxicity.Toxicology,2014.320:p.34-45.


94

205. Slaninova,A.,etal.,Areview:oxidativestressinfishinducedbypesticides.Neuro.EndocrinolLett,2009.30Suppl1:p.2-12.

206. Cattaneo,R.,etal.,ToxicologicalresponsesofCyprinuscarpioexposedtoacommercialformulationcontainingglyphosate.BullEnvironContamToxicol,2011.87(6):p.597-602.

207. Costa,M.J.,etal.,OxidativestressbiomarkersandheartfunctioninbullfrogtadpolesexposedtoRoundupOriginal.Ecotoxicology,2008.17(3):p.153-63.

208. deMenezes,C.C.,etal.,RoundupeffectsonoxidativestressparametersandrecoverypatternofRhamdiaquelen.ArchEnvironContamToxicol,2011.60(4):p.665-71.

209. Ferreira,D.,etal.,AssessmentofoxidativestressinRhamdiaquelenexposedtoagrichemicals.Chemosphere,2010.79(9):p.914-21.

210. Geret,F.,etal.,Effectsoflow-doseexposuretopesticidemixtureonphysiologicalresponsesofthePacificoyster,Crassostreagigas.EnvironToxicol,2013.28(12):p.689-99.

211. Glusczak,L.,etal.,Acuteexposuretoglyphosateherbicideaffectsoxidativeparametersinpiava(Leporinusobtusidens).ArchEnvironContamToxicol,2011.61(4):p.624-30.

212. Guilherme,S.,etal.,DNAdamageinfish(Anguillaanguilla)exposedtoaglyphosate-basedherbicide--elucidationoforgan-specificityandtheroleofoxidativestress.MutatRes,2012.743(1-2):p.1-9.

213. Lushchak,O.V.,etal.,LowtoxicherbicideRoundupinducesmildoxidativestressingoldfishtissues.Chemosphere,2009.76(7):p.932-7.

214. Modesto,K.A.andC.B.Martinez,RoundupcausesoxidativestressinliverandinhibitsacetylcholinesteraseinmuscleandbrainofthefishProchiloduslineatus.Chemosphere,2010.78(3):p.294-9.

215. Ortiz-Ordonez,E.,etal.,EffectofYerbimatherbicideonlipidperoxidation,catalaseactivity,andhistologicaldamageingillsandliverofthefreshwaterfishGoodeaatripinnis.ArchEnvironContamToxicol,2011.61(3):p.443-52.

216. Sinhorin,V.D.,etal.,Effectsoftheacuteexpositiontoglyphosate-basedherbicideonoxidativestressparametersandantioxidantresponsesinahybridAmazonfishsurubim(Pseudoplatystomasp).EcotoxicolEnvironSaf,2014.106:p.181-7.

217. UrenWebster,T.M.,etal.,Effectsofglyphosateanditsformulation,roundup,onreproductioninzebrafish(Daniorerio).EnvironSciTechnol,2014.48(2):p.1271-9.

218. Adamson,P.,etal.,TimetrendsintheregistrationofHodgkinandnon-HodgkinlymphomasinEurope.EurJCancer,2007.43(2):p.391-401.

219. Engels,E.A.,Infectiousagentsascausesofnon-Hodgkinlymphoma.CancerEpidemiolBiomarkersPrev,2007.16(3):p.401-4.

220. Hennessy,B.T.,E.O.Hanrahan,andP.A.Daly,Non-Hodgkinlymphoma:anupdate.LancetOncol,2004.5(6):p.341-53.

221. Muller,A.M.,etal.,Epidemiologyofnon-Hodgkin'slymphoma(NHL):trends,geographicdistribution,andetiology.AnnHematol,2005.84(1):p.1-12.


95

222. Acquavella,J.F.,etal.,GlyphosateBiomonitoringforFarmersandTheirFamilies:ResultsfromtheFarmFamilyExposureStudy.EnvironmentalHealthPerspectives,2003.112(3):p.321-326.

223. Curwin,B.D.,etal.,Urinarypesticideconcentrationsamongchildren,mothersandfatherslivinginfarmandnon-farmhouseholdsiniowa.AnnOccupHyg,2007.51(1):p.53-65.

224. Wester,R.C.,etal.,Glyphosateskinbinding,absorption,residualtissuedistribution,andskindecontamination.Fundam.ApplToxicol,1991.16(4):p.725-732.

225. Wester,R.C.,D.Quan,andH.I.Maibach,Invitropercutaneousabsorptionofmodelcompoundsglyphosateandmalathionfromcottonfabricintoandthroughhumanskin.FoodChemToxicol,1996.34(8):p.731-735.

226. Zouaoui,K.,etal.,DeterminationofglyphosateandAMPAinbloodandurinefromhumans:about13casesofacuteintoxication.ForensicSciInt,2013.226(1-3):p.e20-5.

227. Brewster,D.W.,J.Warren,andW.E.Hopkins,MetabolismofglyphosateinSprague-Dawleyrats:tissuedistribution,identification,andquantitationofglyphosate-derivedmaterialsfollowingasingleoraldose.Fundam.ApplToxicol,1991.17(1):p.43-51.

228. Begley,D.A.,etal.,Findingmousemodelsofhumanlymphomasandleukemia'susingtheJacksonlaboratorymousetumorbiologydatabase.ExpMolPathol,2015.99(3):p.533-6.

229. Hori,M.,etal.,Non-Hodgkinlymphomasofmice.BloodCellsMolDis,2001.27(1):p.217-22.

230. Morse,H.C.,3rd,J.M.Ward,andM.A.Teitell,MousemodelsofhumanBlymphoidneoplasms,inTheLymphoidNeoplasms,I.T.Magrath,Editor.2010,CRCPress:BocaRatan,FL.

231. Pattengale,P.K.andC.R.Taylor,Experimentalmodelsoflymphoproliferativedisease.Themouseasamodelforhumannon-Hodgkin'slymphomasandrelatedleukemias.AmJPathol,1983.113(2):p.237-65.

232. Ward,J.M.,Lymphomasandleukemiasinmice.ExpToxicolPathol,2006.57(5-6):p.377-81.

233. GermanFederalInstituteforOccupationalSafetyandHealth,ProposalforHarmonizedClassificationandLabeling:Glyphosate,F.O.f.Chemicals,Editor.2016:Dortmund,Germany.


96


exhibit 94 · glyphosate acid is a colorless, odorless, crystalline solid. glyphosate is the term...

Documents