Exposure to magnetic fields and childhood acute lymphocytic leukemia in São Paulo, Brazil

Exposure

to

magnetic

fields

and

childhood

acute

lymphocytic

leukemia

in

Sa˜o

Paulo,

Brazil

Victor

Wu¨nsch-Filho

a,

*

,

Daniele

M.

Pelissari

a

,

Fla´vio

E.

Barbieri

b

,

Leonel

Sant’Anna

b

,

Clau´dia

T.

de

Oliveira

c

_,

_Juliana

_F.

_de

_Mata

d

_,

_Luiz

_G.

_Tone

e

_,

_Maria

_Lu´cia

_de

_M.

_Lee

f

_,

Maria

Lydia

M.

de

Andre´a

g

,

Paula

Bruniera

h

,

Sidnei

Epelman

i

,

Vicente

Odone

Filho

k

,

Leeka

Kheifets

j

a

FaculdadedeSau´dePu´blica,UniversidadedeSa˜oPaulo,Brazil

b_{Associac¸a˜oBrasileiradeCompatibilidadeEletromagne´tica,Sa˜oPaulo,Brazil} c_{HospitalAmaralCarvalho,Jau,Brazil}

d

HospitaldoCaˆncerdeBarretos,Brazil

e

FaculdadedeMedicina,UniversidadedeSa˜oPaulo,RibeiraoPreto,Brazil

f

InstitutodeOncologiaPedia´trica,UniversidadeFederaldeSa˜oPaulo,Brazil

g

HospitalInfantilDarcyVargas,Sa˜oPaulo,Brazil

h

SantaCasadeMiserico´rdiadeSa˜oPaulo,Brazil

i_{HospitalSantaMarcelina,Sa˜oPaulo,Brazil}

j_{SchoolofPublicHealth,UniversityofCaliforniaatLosAngeles,USA} k

FaculdadedeMedicina,UniversidadedeSa˜oPaulo,Brazil

1. Introduction

Worldwide leukemia is the most common malignancy in childhood and acute lymphocytic leukemia (ALL) is the most frequenttype,around80%ofallpediatricleukemia[1].Inthelast

twodecades,studiesfromAsia,EuropeandNorthAmericahave reportedincreasedleukemiaincidenceratesinchildren, particu-larlyALL[2–5].Whilemostacceptthattheincreaseinchildhood leukemiaisreal,thereasonsarenotwellunderstood. Although someofthisincreasemaybeduetoimprovementsinthedisease reportinganddiagnosis,thesefactsonlypartlyexplainthetrend. Onehypothesisisthatthistrendmightberelatedtofactorslinked tomoderndevelopmentsinsociety.

Many factorshavebeen suggestedwhichcouldincrease the incidenceofleukemiainchildren,butonlyionizingradiationis recognizedasariskforALLandacute myeloidleukemia[6].In

A R T I C L E I N F O Articlehistory:

Received3December2010 Receivedinrevisedform6May2011 Accepted11May2011

Availableonline15August2011 Keywords:

Electromagneticfields Acutelymphocyticleukemia Children

Case–controlstudy Brazil

A B S T R A C T

Background:Epidemiologicalstudieshaveidentifiedincreasedrisksofleukemiainchildrenlivingnear powerlinesandexposedtorelativelyhighlevelsofmagneticfields.Resultshave beenremarkably consistent,butthereisstillnoexplanationforthisincrease.Inthisstudyweevaluatedtheeffectof60Hz magneticfieldsonacutelymphocyticleukemia(ALL)intheStateofSa˜oPaulo,Brazil.Methods:Thiscase– controlstudyincludedALLcases(n=162)recruitedfromeighthospitalsbetweenJanuary2003and February2009.Controls(n=565)matchedongender,age,andcityofbirthwereselectedfromtheSa˜o PauloBirthRegistry.Exposureto extremelylowfrequencymagneticfields(ELFMF)was basedon measurementsinsidehomeanddistancetopowerlines.Results:For24hmeasurementsinchildren rooms,levelsofELFMFequaltoorgreaterthan0.3microtesla(mT),comparedtochildrenexposedto levelsbelow0.1mTshowednoincreasedriskofALL(oddsratio[OR]1.09;95%confidenceinterval[95% CI]0.33–3.61).Whenonlynighttimemeasurementswereconsidered,arisk(OR1.52;95%CI0.46–5.01) wasobserved.Childrenlivingwithin200mofpowerlinespresentedanincreasedriskofALL(OR1.67; 95%CI0.49–5.75),comparedtochildrenlivingat600mormoreofpowerlines.Forthoselivingwithin 50mofpowerlinestheORwas3.57(95%CI0.41–31.44).Conclusions:Eventhoughourresultsare consistentwiththesmallrisksreportedinotherstudiesonELFMFandleukemiainchildren,overallour resultsdonotprovidesupportforanassociationbetweenmagneticfieldsandchildhoodleukemia,but smallnumbersandlikelybiasesweakenthestrengthofthisconclusion.

ß2011ElsevierLtd.

*Correspondingauthorat:DepartamentodeEpidemiologia,FaculdadedeSau´de Pu´blica,UniversidadedeSa˜oPaulo,Av.Dr.Arnaldo,715,Sa˜oPaulo,SP,Brazil. Tel.:+551130617764;fax:+551130617926.

E-mailaddress:wunsch@usp.br(V.Wu¨nsch-Filho).

ContentslistsavailableatScienceDirect

Cancer

Epidemiology

The

International

Journal

of

Cancer

Epidemiology,

Detection,

and

Prevention

j o urn a lhom e pa g e :ww w . ca nc e re pi d e mi ol o gy . ne t

1877-7821 ß2011ElsevierLtd.

doi:10.1016/j.canep.2011.05.008

Open access under the Elsevier OA license.

1979,Wertheimerand Leeper [7]reported an increasedriskof leukemiaand braintumors inchildren living nearpower lines. Sincethen,numerous epidemiologicalstudiesof extremelylow frequency magnetic fields (ELF MF) have been performed in variouscountries.Ofnotearetwopooledanalyses,oneconducted byGreenlandetal.[8],whichaggregated12case–controlstudies, andreportedoddsratio(OR)of1.7(95%confidenceinterval[95% CI] 1.2–2.3) for those exposed to levels higher than 0.3

m

T comparedwithlevelslowerthan0.1

m

T;andanotherconducted byAhlbometal.[9]withninecase–controlstudiesandanORof2.0 (95%CI1.3–3.1)inthehighestexposurecategory(levelsequaltoor higherthan0.4

m

T).

Basedon theevidence,particularlyresultsfrombothpooled analyses,theInternationalAgencyforResearchonCancer(IARC) classifiedELFMFasclass2B,oraspossiblycarcinogenictohumans

[10]. In 2007, the World Health Organization Environmental HealthCriteriaconfirmedtheIARCclassificationofthe epidemio-logicalevidenceforchildhoodleukemia[11].

WhilestudiesontheeffectofELFMFexposureonchildhood leukemiahavebeenconductedinmanycountries,nonehavebeen doneinSouthAmerica.In thisstudyweevaluatedtheeffectof exposure to60Hz magneticfields on theoccurrence of ALLin childrenlivingintheStateofSa˜oPaulo,Brazil.

2. Materialsandmethods

This case–control study included incident cases of ALL diagnosed from January 1st, 2003to February 20th, 2009 and bornonJanuary1st,2000orlater.Inordertobeincludedinthe study,thechildwithALLhadtohavelivedintheStateofSa˜oPaulo. AmaximumperiodofsixmonthsresidenceoutsidetheStateofSa˜o Paulowasallowed.Visitstothehomesofcasesandcontrolsfor measurementsof ELFMF were performed fromFebruary 15th, 2006toMarch25th,2009.

2.1. Selectionofcases

Caseswererecruitedfromeighthospitals,whichconcentrate onthecareofchildrenwithcancerandcovermorethan50%ofall leukemiacasesinchildrenintheStateofSa˜oPaulo.ALLdiagnosis was based on clinical patterns, results from blood analysis, immunophenotyping,andwhenavailable,cytogeneticand molec-ularbiologyanalysis.

ThemotherofthechildwithALL(oralternativelythefatheror otherresponsibleperson)wasinterviewedinthehospitalusinga structuredquestionnaire.Ineachhospital,amemberoftheclinical staff was responsible for conducting the interviews. Cases that moved post diagnosis were excluded from the study as it was impossibletoconductELFMFmeasurementsinthehomeswerethey hadlivedpriortodiagnosis,i.e.,duringtherelevantetiologicperiod. Atotalof248eligiblecaseswerereportedfromthehospitals, but55(22.2%)hadmovedfromtheresidenceofALLdiagnosisand 14(5.6%)couldnotbelocatedandthereforecouldnotbevisitedfor measurements.Fromthe179remainedcases,17(9.5%) refused participateinthestudy.

2.2. Selectionofcontrols

FromtheStateofSa˜oPauloBirth Registrywetriedtoobtain fourcontrolsmatchedtocaseongender,age(25%foragebelow4 years; 1yearforageabove4 years)andcity ofbirth,whenno controlswere available in case’s city of residence, controls were selectedfromadjacentmunicipalities,butalwayswithintheStateof Sa˜oPaulo.

Twentyeligiblepotentialcontrols,ifavailable,werelistedfor eachcase fromtheStateofSa˜oPauloBirth Registry.Then, four

weredrawnatrandomforhomevisit.Theinterviewersreceived thelistofcontrolsalreadyintheordertheyshouldbeinterviewed. Thelistofcontrolswaspreparedatrandombyafieldcoordinator whowasunawareoftheaddresslocationandifitwasnearapower lineornot.Inshort,neitherthepersonwhopreparedthelistof controls, northeinterviewer knew thelocation ofthecontrols addresspriortoselection.Ifsomeoffourcontrolswerenotfound after two visits,replacementcontrols wereselectedat random fromthe list.Thisprocess continued,whenpossible,until four controlswereselectedforeachcase.

From the 2500 potential controls selected for visits, we ascertainedthat1108(44.3%)hadmoved,747(29.9%)wereabsent, andthree(0.1%)wereexcludedfromtheanalysisastheyhadlived outsidetheStateofSa˜oPauloformorethansixmonths.Fromthe 642remainedcontrolscontacted,77(12%)refusedtoparticipatein thestudy.Asaresult,atotalof565controlswereinterviewedand ELFMFexposuremeasurementsweremadeintheirhomes.

Forthe162participatingcases,thenumberofcontrolspercase variedfromzerotofour.For117cases(72.2%)fourcontrolswere obtained,for22(13.6%)three,for12(7.4%)two,andforsevencases (4.3%)onlyonecontrolwasavailable.Forfourcases(2.5%)controls couldnotbefound.

Eachhouseholdwasvisitedtointerviewthemother(orfather orotherresponsibleperson)usingaquestionnairesimilartothe oneusedforcases.Childrenselectedaspotentialcontrols,butwith aprevioushistoryofleukemiaorothercancerswereexcludedfrom thestudy.

2.3. AssessmentofexposuretoELFMF

Twoapproaches wereusedtoevaluateexposure toELFMF. First,ELFMFexposuremeasurementsweremadeinsidecaseand controlhomesusingaportableEMDEX-IIdosimeter(40–800Hz, Enertech, Campbell, CA,USA).Measurements of3min duration weretakenoutsidethefrontdoorandineachroomofthehouse. Subsequently,thedosimeterwasleftunderthechild’sbedfor24h. Dosimeters were checkedevery week and calibrated every six months.The24hinhome ELFMFwerecategorizedforanalysis intolessthan0.1

m

T,0.1

m

Ttolessthan0.3

m

T,andequaltoor greaterthan0.3

m

T,alsolevelsequaltogreaterthan0.4

m

Twere analyzed.

Inthesecondapproach,thedistancewascalculatedbetween caseandcontrolhouseholdsandtheclosest88,138,230,345,or 440kilovolt(kV)powerlines.Thisprocedurewasrestrictedtothe subsetofcasesandcontrolslivingintheMetropolitanRegionof Sa˜oPaulo(MRSP)becausethiswastheonlyregionintheStateof Sa˜oPaulowhereelectric gridmapswereavailable.MRSPhasa networkof 572km ofpower lines.AGlobalPositioningSystem (GPS) was used to spatially locate each house address. GPS measurementsweretransferredtoacomputerizeddatabaseand thendistancesofthehouseholdtopowertransmissionlineswere calculated usingMapInfosoftware forgeographical information systems(GIS).WeusedthedistancestratificationbyDraperetal.

[12]:distanceofhouseholdsfrompowerlineswascategorizedinto below100m,100tolessthan200,200tolessthan600,andequal toorgreaterthan600m.We alsoexamined theriskofALLfor thoselivinglessthan50mfrompowerlines;aswellastheriskof thoselivingnear88kVpowerlinesandthoselivingnearhigher voltagepowerlines,asthemagneticfieldsingeneralarehigherin proximitytohighervoltages.

2.4. Comparisonbetweencasesandcontrolsparticipatingandnot participatinginthestudy

Table1comparescasesandcontrolsnotincludedforanyreason withthoseincludedinthestudy.Nodifferenceswereobservedin

gender; however, not participating children were older than participatingchildren.Therewerealsosomeregiondifferencesof residencebetweenincludedandnon-includedcasesandcontrols; ahigherpercentageofparticipatingcasesandlowerpercentageof participatingcontrols were from the MRSP.Higher portions of casesand controls includedin thestudy lived closerto power transmission lines than those not included. While front door measurements did not differ between participating and not participatingcases,ahigherpercentageofincludedcontrolshad lowerfrontdoormeasurements(ELFMFexposureslower0.1

m

T) thannon-includedcontrols.

2.5. Statisticalanalysis

We performed conditionalandunconditionallogistic regres-sionanalysis,astheresultswereequivalentwepresentedonly results from the conditional logistic regression analysis. We calculatedtheoddsratiosandrespective 95%CItoevaluatethe effect of ELF MF exposure and ALL incidence. Analysis was performedusingSAS@_andSPSS@_{software.Potentialconfounding}

variables,suchasage,gender, race(white/non-white),mobility, schoolingoftheinterviewed,Downsyndrome,child’shistoryof influenza,child’sattendanceindaycareorschool,ageofmotherat interview,mother’shistoryofworkinagriculture,and mother’s history of smoking tobacco and alcohol consumption were examined by adding covariates to the conditional logistic regressionmodels.

3. Results

Table2showssomeadditionalcharacteristicsofthe162cases and565controlsincludedintheanalysis.Themajorityofcasesand controlswasclassifiedaswhite.TheprevalenceofDownsyndrome washigherin cases.Mostinterviewswerecarriedoutwiththe

child’smother.51.3%casesand60.4%controlshadnineormore yearsofschooling.Consideringstudylogistics,andasexpected,the rateofchildrenwhohadmovedwashigherforcases(42.0%)than controls(5.8%).

Mean24hELFMFexposurelevelsmeasuredunderchildren’s bedswerecomparableforcases(0.104

m

T,standarddeviation[SD] 0.190

m

T)andcontrols(0.113,SD0.130

m

T).Noincreasedriskwas observedinthecrudeandadjustedanalysisforthoseexposedto

Table1

Casesandcontrolsnotincludedandincludedinthestudycomparedbydemographicalvariables,livingregion,distanceofhometopowerlineandlevelsofmagneticfields exposure.

Cases Controls

Notincludeda _{Included pValue Notincluded}a _{Included pValue}

n=113 % n=162 % n=1935 % n=565 % Sex Male 59 52.2 86 53.1 0.886 963 49.9 297 52.6 0.274 Female 54 47.8 76 46.9 965 50.1 268 47.4 Ageatinterview <2years 11 9.8 41 25.3 <0.001 239 12.4 96 17.0 0.019 2to<4years 33 29.5 53 32.7 495 25.8 154 27.3 4to<6years 44 39.3 57 35.2 674 35.1 183 32.4 6ormoreyears 24 21.4 11 6.8 514 26.7 132 23.4 Livingregion MRSPb 53 49.5 121 74.7 <0.001 1591 82.3 418 74.0 <0.001 Otherregionsc 37 34.6 41 25.3 342 17.7 147 26.0 Otherstates 17 15.9

Distanceofhometopowerline(m)d

600 51 96.2 86 71.1 <0.001 1181 74.2 281 67.3 0.025 200–599 1 1.9 20 16.5 270 17.0 88 21.1 100–199 0 10 8.3 78 4.9 25 6.0 <100 0 5 4.1 59 3.7 24 5.7 Missing 1 1.9 3 0.2 Externalmeasurementse <0.1mT 20 64.5 96 59.3 0.675 885 47.7 315 55.9 0.002 0.1to<0.3mT 7 22.6 49 30.2 683 36.8 171 30.4 0.3mT 4 12.9 17 10.5 289 15.5 77 13.7 a

Allreasons.Seetext.

b_{MetropolitanRegionofSa˜oPaulo.} c_{OtherregionsofSa˜oPauloState.} d

OnlycasesandcontrolsfromtheMetropolitanRegionofSa˜oPaulo(MRSP).

e

Externalmeasurementsconductedatthehouseholdfrontdoor.

Table2

Characteristicsofparticipatingcasesandcontrol.

Variables Cases Controls

n=162 % n=565 %

Race

White 102 63.0 372 65.8

Non-white 59 36.4 189 33.5

Missing 1 0.6 4 0.7

Numberofbrothersandsisters

None 47 29.0 127 22.5 Only1 50 30.9 227 40.2 2 64 39.5 211 37.3 Missing 1 0.6 0 – Downsyndrome Yes 6 3.7 4 0.7 No 135 83.3 549 97.2 Missing 21 13.0 12 2.1 Interviewedperson Mother 153 94.4 441 78.1 Father 5 3.1 39 6.9 Other 4 2.5 85 15.0

Schoolingofinterviewedperson(years)

Never 5 3.1 9 1.6

<4years 20 12.3 83 14.7

5–8years 54 33.3 132 23.4

9–11years 55 34.0 162 28.7

levels equal to or higher than 0.3

m

T (Table 3). When only consideringmeasurementsatnightanincreasedadjustedriskwas observed for those exposed at levels equal to or greater than 0.3

m

T;andthisriskwasevenhigherinanalysisthatonlyincluded cases and controls who had never moved (Table 3). However, estimates were imprecise and included a null value and risk deficits were observed in the intermediate category, with exposuresfrom0.1to<0.3

m

T.Forthoseexposedtolevelsequal togreaterthan0.4

m

T,theadjustedORwere1.01(95%CI0.23– 4.42)and0.18(95%CI0.1–3.20),respectivelyfor24hexposureand nightexposure(datanotshown).

Onlyfivecasesand 24controlslivedwithin100mofpower lines.Thesenumbersdroppedtofourcasesand11controlswhen thedistancewaswithin50m.IncreasedriskofALLwasobserved forchildrenlivingwithin100mfrompowerlinesandbetween100 and200mintheMRSP(Table4).Whenanalysiswasrestrictedto casesandcontrolswhohadhadnevermoved,theriskwashigher for children living between 100 and 200m, but remained unchanged for those closer than 100m from power lines. For thosewithin50mofpowerlinestheORwas3.57(95%CI0.41– 31.44;datanotshown),butbasedononlyfourcases.Ariskdeficit wasobservedatdistanceof200to599m.Mostcases(72.6%)and

controls(81.2%)hadlivedcloseto88kVlines.Analysisrestricted tothisgroupshowedariskforthoseliving100–200mfrompower lines(OR4.04;95%CI0.70–23.38)andariskdeficitforthosewithin 100mfrompowerlines(OR0.39;95%CI0.01–15.85).Forthose whohadlivedcloseto>88kVpowerlines,riskcalculationwasnot possibleduetothesmallnumberofobservations(datanotshown). 4. Discussion

WedidnotobserveincreasedriskofALLforexposuresequalto or greater than 0.3

m

T estimated by 24h measurements in children’s bedrooms.However, when analysiswas restrictedto nighttimebedroommeasurementsanincreasedriskwasobserved atthesameexposurelevel.Whenrestrictedtocasesandcontrols whonevermoved,nighttimeexposureriskwasevenhigher.This resultofriskofleukemiainchildrenexposedtohigherlevelsofELF MFduringnightisanalogoustothatreportedinGermany,butnot inCanada,UnitedKingdom,orUnitedStates[13].

WeobservedincreasedriskofALLamongchildrenlivingwithin 50mfrompower lines,fordistances ofless than100mand at distancesfrom100to200m.Severalformerstudieshavefounda relationshipbetweendistanceofhouseholdtopowerlinesandrisk

Table3

Risks(crudeandadjusted)ofacutelymphocyticleukemiaaccordingtoexposuretomagneticfieldsmeasurements(24handnightexposure)insidethehousebydosimeter, Sa˜oPauloState.

mT Cases Controls CrudeORa

(24hexposure)b (95%CI)c AdjustedORd (24hexposure)b (95%CI)c AdjustedORd (nightexposure)e (95%CI)c n=162 % n=565 %

Allcasesandcontrols

<0.1 113 69.8 394 69.7 1.00 1.00 1,00

0.1to<0.3 38 23.5 137 24.2 0.97(0.64–1.47) 0.75(0.36–1.55) 0.48(0.21–1.11)

0.3 11 6.7 34 6.0 1.13(0.55–2.30) 1.09(0.33–3.61) 1.52(0.46–5.01)

Onlycasesandcontrolsthatnevermovedofhouseholde

<0.1 68 72.3 373 70.0 1.00 1.00 1,00

0.1to<0.3 22 23.4 125 23.5 0.97(0.57–1.63) 0.98(0.39–2.49) 0.50(0.17–1.45)

0.3 4 4.3 35 6.6 0.63(0.22–1.82) 0.98(0.18–5.48) 1.72(0.30–9.87)

a_Oddsratio.

b _{Averageofexposuremeasurementsfor24h.} c

95%confidenceinterval.

d

Oddsratioadjustedinthelogisticregressionmodelaccordingtoage,gender,race(white/non-white);mobility(nevermoved/movedatleastonce);educationof interviewedperson(yearsofformalschooling);daycare(attended/notattended);Downsyndrome(no/yes);influenzahistory(no/yes);ageofthemotheratinterview; mother’shistoryofagriculturalworkduringlifetime(never/ever);mother’shistoryofsmoking(never,former,current);mother’shistoryofalcoholconsumption(never, former,current).

e_{Averageofexposuremeasurementsatnight:from8.00PMto8:00AM.Inthisanalysiswasexcludedthevariablemobility.}

Table4

Risks(crudeandadjusted)ofacutelymphocyticleukemiabydistanceofhouseholdtotheclosestpowertransmissionline,MetropolitanRegionofSa˜oPaulo.

Distance Cases Controls CrudeORa_(95%CI)b _AdjustedORc_(95%CI)b

n=121 % n=418 %

Allcasesandcontrols

600 86 71.1 281 67.2 1.00 1.00

200to<600 20 16.5 88 21.1 0.74(0.43–1.28) 0.69(0.28–1.71)

100to<200 10 8.3 25 6.0 1.31(0.60–2.83) 1.67(0.49–5.75)

<100 5 4.1 24 5.7 0.68(0.25–1.84) 1.54(0.26–9.12)

Onlycasesandcontrolsthatnevermovedofhouseholdd

600 53 72.6 266 67.2 1.00 1.00 200to<600 9 12.3 85 21.5 0.53(0.25–1.12) 0.91(0.25–3.25) 100to<200 9 12.3 22 5.6 2.05(0.90–4.71) 3.68(0.68–19.82) <100 2 2.8 23 5.7 0.44(0.10–1.91) 1.52(0.11–21.24) a Oddsratio. b _{95%confidenceinterval.}

c_{Oddsratioadjustedinthelogisticregressionmodelaccordingtoage,gender,race(white/non-white);mobility(nevermoved/movedatleastonce);educationof}

interviewedperson(yearsofformalschooling);daycare(attended/notattended);Downsyndrome(no/yes);influenzahistory(no/yes);ageofthemotheratinterview; mother’shistoryofagriculturalworkduringlifetime(never/ever);mother’shistoryofsmoking(never,former,current);mother’shistoryofalcoholconsumption(never, former,current).

d

ofchildhoodleukemia.Particularly,Draperetal.[12]havefound increasedriskofleukemiaamongchildrenlivingwithin200mof power lines, however, and without a cogent explanation, an increasedriskpersistedatdistancesbetween200and600m,far beyondthedistanceswherenohigherELFMFareexpected.The majorityofchildren in theMRSPlivednear 88kVpower lines. Whenwerestrictedtheanalysistothissubgroupofchildrenthe riskforthoselivingbetween100and200mincreased substan-tially,howeverfieldsfromlowvoltagelinesareexpectedtobeless thanfieldsatcorrespondingdistancesfromhighervoltagelines.In theMRSP,the88kVpowerlinesareusuallylocatedintheareas withhighpopulationdensity.Thustheobservedriskcouldbedue tootherfactors,suchashigherexposuretobenzenefromheavy traffic.

Animportant,butunexplained,observationisasubstantialrisk deficitintheintermediatecategoryin bothapproachesused to evaluateexposuretomagneticfields,directmeasurementsinside thehouseanddistanceofhouseholdtopowerlines.Similardiphas been observed for calculated (but not for measured fields) in previousmeta-analysis[9,14].

EpidemiologicalstudiesofELFMFandleukemiainchildrenare difficulttobe design, conduct,and interpret. There are several reasonsforthis–themostproblematicisvalidityandprecisionin assessingexposuretoELFMF,astheyvaryintimeandspaceand their assessment is difficult, especially in retrospective case– controlstudies.However,therarityofchildhoodleukemiamakes onlyretrospectivestudiesfeasible.Theprevalenceofexposureto ELFMFatlevelsequaltoorgreaterthan0.3

m

Tinthepopulationis verylow, makingit difficultto obtainreliableestimates of the potentialinfluenceofELFMFathigherlevels.

BesidesthedifficultyofevaluatingELFMF,manystudieson the risk of childhood leukemia and ELF MF were prone to selectionbias[15].Selectionbiasoccurswhentheparticipation ofcasesandcontrols aredifferentandtheparticipationrateis correlatedwithexposure[16].Dependingonthetypeofstrategy adopted for the evaluation of ELF MF, studies have reported participationratesrangingfrom33%to79%.Whenavailablethe participationrate, in general, is higher in cases than controls

[17].Asoverallenrollmentrateswerelowthepotentialforbias in our study is substantial. Inaddition, participating and not participating children in our study differed on number of parametersincludingsomecrudeestimatesofexposure,suchas distanceofthehousetopowerlineandELFMFmeasurementsin thefrontdoor.

Weencounteredsomedifficultiesspecificallyrelatedtostudies conducted in Brazil. Due to a lack of older well organized populationrecords,suchasbirthregisters,wewererestrictedto thefewyearsforwhichdatawasavailablethusleadingtoasmall samplesize.Furthermore,wewereonlyabletoincludechildren aged eight or younger, because computerized records birth certificaterecords,usedforcontrolselection,wereonlyavailable from2000forward.Asaresult,thecasesinourstudywerealittle younger than controls and differed from other studies, which includechildren0–14yearsofage.

Inaddition,thereareseveralparticularitiesthatmighthaveled tobias.ItiscommoninBrazilforfamiliesofchildrenwithleukemia tomoveclosetothetreatinghospital,andsubjectswhomoved afterdiagnosiswerenotincludedasitwaslogisticallyinfeasibleto conductmeasurementsinthehomeswheretheyhadlivedpriorto diagnosis.Toaddressthisproblem,welimitedsomeoftheanalysis toastableandpresumablymorecomparablesubset.Weobserved higherriskswhenwe restrictedanalysistochildren (casesand controls)wholivedinasingleresidence.

Finally, while adjustments for various confounders did not makeadifferenceinmostofthepreviousstudies,theadjustment forseveralfactorsinourstudyhadaninfluenceonriskestimates.

Totheextentthatevaluationofthesefactorsispronetobias;these adjustmentsarelikelytobeincomplete.

Inconclusion,wedidnotobservedanincreasedriskofALLfor childrenwithELFMFexposuresequalorabove0.3

m

Torabove 0.4

m

T comparedtothose exposedtolevels lowerthan0.1

m

T. Increasedriskswereobservedinsomesubgroup,butresultswere inconsistent,impreciseandincludedanullvalue.Whileourresults areconsistentwiththesmallrisksreportedinotherstudiesonELF MFandleukemiainchildren,overalldonotprovidesupportforan associationbetweenmagneticfieldsandchildhoodleukemia,but small numbers and likely biases weaken the strength of this conclusion.

Conflictofintereststatement

Ideclaretohavenoconflictofinterestrelatedtothispaper. Acknowledgments

This work was conducted under the EMF Project-SP, P&D project, with resources allocated by the Brazilian Electricity RegulatoryAgency(0390-041/2004 ANNEL),coordinatedbythe Brazilian Association for Electromagnetic Compatibility (ABRI-CEM),andperformedattheUniversityofSa˜oPauloSchoolofPublic Health.TheauthorsaregratefultoMonicaLaPorteTeixeiraand Rosa Maria Vieira Freitas of theFundac¸a˜o Sistema Estadual de Ana´lisedeDados(SEADE),whocontributedtoselectionofcontrols from Sa˜o Paulo Birth Registry; Aparecida Nata´lia Rodrigues, MoˆnicaMagalha˜esLeite,andRosemairyInamine,whocoordinated theselectionof casesand controlsand theinterviews;and the team ofinterviewers, whointerviewed casesand controls;and AndersondaCostaformanagementofthedatabank.

TheauthorsalsothanktheengineersHamiltonMossdeSouza and Paulo Roberto Gonc¸alves de Oliveira from the Centro de PesquisasdeEnergiaEle´trica(CEPEL)ofRiodeJaneiro;MarioLeite PereiraFilhofromInstitutodePesquisasTecnolo´gicasdoEstadode Sa˜o Paulo (IPT) for their contributions during magnetic field measurementsprocedures;andMiriamReginadeSouzafromthe DepartamentodeMedicinaPreventivadaFaculdadedeMedicina daUniversidade deSa˜oPaulofor hercolaborationin preparing mapsoftheelectricgridforanalysis.

References

[1]BrennerH,CoeberghJW,ParkinDM,IzarrugazaI,ClavelJ,ArndtV,etal. Up-to-datemonitoringofchildhoodcancerlong-termsurvivalinEurope:leukemia andlymphomas.AnnOncol2007;18:1569–77.

[2]CoeberghJWW,ReedijkAMJ,deVriesE,MartosC,JakasZ,Steliarova-Foucher E,etal.Leukaemiaincidenceandsurvivalinchildrenandadolescents in Europeduring 1978–1997:reportfromtheAutomatedChildhoodCancer InformationSystemproject.EurJCancer2006;42:2019–36.

[3]ShahA,ColemanMP.Increasingincidenceofchildhoodleukaemia:a contro-versyre-examined.BrJCancer2007;97:1009–12.

[4]LinaberyAM,RossJA.TrendsinchildhoodcancerincidenceintheU.S.(1992– 2004).AmCancerSoc2008;112:416–32.

[5]SwaminathanR,RamaR,ShantaV.ChildhoodcancersinChennaiIndia,1990– 2001:incidenceandsurvival.IntJCancer2008;122:2607–11.

[6]BelsonM,KingsleyB,HolmesA.Riskfactorsforacuteleukemiainchildren:a review.EnvironHealthPerspect2007;115:138–45.

[7]WertheimerN,LeeperE.Electricalwiringconfigurationandchildhoodcancer. AmJEpidemiol1979;109:273–84.

[8]GreenlandS,SheppardAR,KauneWT,PooleC,KelshMA.Apooledanalysisof magneticfields,wirecodes,andchildhoodleukemia.Epidemiology2000;11: 624–34.

[9]AhlbomA,DayN,FeychtingM,RomanE,SkinnerJ,DockertyJ,etal.Apooled analysisofmagneticfieldsandchildhoodleukaemia.BrJCancer2000;83: 692–8.

[10]IARCWorkingGroupontheEvaluationoftheCarcinogenicRiskstoHumans. Staticandextremelylow-frequencyelectricandmagneticfields,vol.80.Lyon: IARCPress,2002.

[11]WorldHealthOrganization. Extremelylowfrequencyfields,vol.238.Spain: WHO,2007.

[12]DraperG,VincentT,KrollME,SwansonJ.Childhoodcancerinrelationto distancefromhighvoltagepowerlinesinEnglandandWales:case–control study.BJM2005;330:1290–3.

[13]Schu¨zJ,SvendsenAL,LinetMS,McBrideML,RomanE,FeychtingM,etal. Nighttimeexposuretoelectromagneticfieldsandchildhoodleukemia:an extendedpooledanalysis.AmJEpidemiol2007;166:263–9.

[14]KheifetsL,AhlbomA,CrespiCM,DraperG,HagiharaJ,LowenthalRM,etal. Pooledanalysisofrecentstudiesonmagneticfieldsandchildhoodleukaemia. BrJCancer2010;103:1128–35.

[15]KheifetsL,ShimkhadaR.ChildhoodleukemiaandEMF:reviewofthe epide-miologicevidence.Bioelectromagnetics2005;7:S51–9.

[16]BreslowNE,DayNE.Statisticalmethodsincancerresearch,volI:theanalysis ofcase–controlstudies.IARCScientificPublicationNo.32.Lyon:IARCPress, 1980.p.5–338.

[17]MezeiG, Kheifets L. Selectionbias andits implicationsfor case–control studies: a casecontrol studyof magnetic field exposureand childhood leukemia.IntJEpidemiol2006;35:397–406.