Review
Reproductive and developmental toxicity of formaldehyde: A systematic review
Anh Duong
a, Craig Steinmaus
a,b, Cliona M. McHale
a, Charles P. Vaughan
c, Luoping Zhang
a,*
aSchoolofPublicHealth,UniversityofCalifornia,Berkeley,CA94720,UnitedStates
bOfficeofEnvironmentalHealthHazardAssessment,CaliforniaEnvironmentalProtectionAgency,Oakland,CA94612,UnitedStates
cGlobalHealthSciences,UniversityofCalifornia,SanFrancisco,CA94143,UnitedStates
Contents
1. Introduction... 119
1.1. Previousreviewsofformaldehydereproductiveanddevelopmentaltoxicity... 119
1.2. Systematicapproachofthecurrentreview... 120
2. Humanpopulationstudies... 120
2.1. Reproductivetoxicityinhumans... 121
2.2. Developmentaltoxicityinhumans... 121
2.2.1. Spontaneousabortion... 121
2.2.2. Congenitalanomalies ... 123
2.2.3. Lowbirthweight... 124 MutationResearch728(2011)118–138
ARTICLE INFO Articlehistory:
Received18April2011
Receivedinrevisedform9July2011 Accepted9July2011
Availableonline20July2011 Keywords:
Formaldehyde Teratogenicity Pregnancy Meta-analysis Human Animal
ABSTRACT
Formaldehyde,therecentlyclassifiedcarcinogenandubiquitousenvironmentalcontaminant,haslong beensuspectedofcausingadversereproductiveanddevelopmentaleffects,butpreviousreviewswere inconclusive,dueinpart,tolimitationsinthedesignofmanyofthehumanpopulationstudies.Inthe currentreview,wesystematicallyevaluatedevidenceofanassociationbetweenformaldehydeexposure andadversereproductiveanddevelopmentaleffects,inhumanpopulationsandinvivoanimalstudies,in the peer-reviewed literature. The mostly retrospective human studies provided evidence of an association of maternal exposure with adverse reproductive and developmental effects. Further assessmentofthisassociationbymeta-analysisrevealedanincreasedriskofspontaneousabortion (1.76,95%CI1.20–2.59,p=0.002)andofalladversepregnancyoutcomescombined(1.54,95%CI1.27–
1.88, p<0.001), in formaldehyde-exposed women, although differential recall, selection bias, or confoundingcannotberuledout.Evaluationoftheanimalstudiesincludingallroutesofexposure,doses anddosingregimensstudied,suggested positiveassociationsbetween formaldehydeexposureand reproductive toxicity, mostly in males. Potential mechanisms underlying formaldehyde-induced reproductiveanddevelopmentaltoxicities,includingchromosomeandDNAdamage(genotoxicity), oxidative stress, altered level and/or function of enzymes, hormones and proteins, apoptosis, toxicogenomicand epigenomiceffects(suchasDNAmethylation),wereidentified.Toclarifythese associations, well-designed molecular epidemiologic studies, that include quantitative exposure assessmentanddiminishconfoundingfactors,shouldexaminebothreproductiveanddevelopmental outcomesassociated with exposure inmales and females.Together withmechanistic and animal studies,thiswillallowustobetterunderstandthesystemiceffectofformaldehydeexposure.
ß2011ElsevierB.V.Allrightsreserved.
Abbreviations:ACTH,adrenocorticotropinhormone;CI,confidenceInterval;CRH,corticotropinreleasinghormone;DPC,DNAproteincrosslinking;EPA,environmental protectionagency;FDR,fecundabilitydensityratio;GSH,glutathione;GSH-PX,glutathioneperoxidase;HPA,hypothalamus–pituitary–adrenal;IARC,InternationalAgency forResearchonCancer;IRIS,IntegratedRiskInformationSystem;i.g.,intragastricinjection;i.m.,intramuscularinjection;i.p.,intraperitonealinjection;i.v.,intravenous injection;ir,immunoreactive;LDH,lactatedehydrogenase;MDA,malondialehyde;MN,micronuclei;MOA,mechanismofaction;NIOSH,NationalInstituteforOccupational SafetyandHealth;NTP,NationalToxicologyProgram;OR,oddsratio;OSHA,OccupationalSafetyandHealthAdministration;ppm,partspermillion;RoC,ReportonCancer;
ROS,reactiveoxygenspecies;RR,relativerisk;SAB,spontaneousabortion;SCE,sisterchromatidexchange;SDH,succinatedehydrogenase;SE,standarderror;SOD, superoxidedismutase.
*Correspondingauthorat:SchoolofPublicHealth,UniversityofCaliforniaatBerkeley,237HildebrandHall,Berkeley,CA94720-7356,UnitedStates.Tel.:+15106435189.
E-mailaddress:luoping@berkeley.edu(L.Zhang).
ContentslistsavailableatScienceDirect
Mutation Research/Reviews in Mutation Research
j our na l ho me p a ge : w ww . e l se v i e r . com / l oc a te / r e v i e ws mr C om mu ni t ya dd re ss : ww w . e l se v i e r. c om / l oca t e / m ut re s
1383-5742/$–seefrontmatterß2011ElsevierB.V.Allrightsreserved.
doi:10.1016/j.mrrev.2011.07.003
2.2.4. Prematurebirth... 124
2.2.5. NIOSHanalysisofcombinedadversebirthoutcomes... 124
2.3. Limitationsofthehumanstudies ... 124
3. Updatedmeta-analysis... 124
3.1. PreviousresultsfromCollinsetal. ... 124
3.2. Uniqueapproachinthepresentanalysis... 125
3.3. Meta-analysismethods... 125
3.4. Results ... 126
3.5. Discussion... 127
4. Experimentalanimalstudies ... 128
4.1. Exposureroutesrelevanttohuman ... 128
4.2. Reproductivetoxicity... 131
4.2.1. Rats ... 131
4.2.2. Mice... 131
4.2.3. Otheranimals ... 132
4.3. Developmentaltoxicity... 132
4.3.1. Rats ... 132
4.3.2. Mice... 132
4.3.3. Otheranimals ... 132
4.3.4. Post-natalexposureanddevelopmentaltoxicity... 132
4.4. Exvivoandinvitroanimalstudies... 132
4.5. Keyfindings ... 133
5. Potentialmechanismsofaction... 133
5.1. ChromosomaldamageandDNAlesions... 133
5.2. Oxidativestress... 133
5.3. OtherpossibleMOAs ... 134
5.3.1. Dehydrogenases ... 134
5.3.2. Heatshockproteins... 134
5.3.3. Apoptosis... 134
5.3.4. Epigeneticalterations... 134
5.3.5. Sexhormones ... 134
5.3.6. Hypothalamus–pituitary–adrenalglandaxis... 134
5.4. Comparisonwithotherreproductivetoxicants... 135
6. Currentresearchgapsandfuturedirections... 135
7. Conclusion... 136
Acknowledgements... 136
References... 136
1. Introduction
Withmorethan46billionpoundsproducedworldwideannually [1], most of which is widely used in the construction, textile, furniture,medical,chemical,andpharmaceuticalindustries,form- aldehydeheavilyimpactstheeverydayconsumer.Itisproduced endogenously in all living organisms, including humans, but exposuretoubiquitous exogenous sourcesindoors,outdoors, at work,inresidences,infoodandmedicine,posesasignificantthreat topublic health[2].Exposedpopulationsincludenotonlyadult workers, who are exposed occupationally, but also the elderly, childbearing women, and young children. Emerging evidence supports an association between formaldehyde exposure and multipleadversehealtheffects[2].Itisincreasinglybeingaccepted byInternationalAgencyforResearchonCancer(IARC)[1,3],theUS NationalToxicologyProgram(NTP)(12thRoC,ReportonCancer)[4], and the US Environmental Protection Agency (EPA) [5], that formaldehydeisa humancarcinogen.Althoughit haslongbeen suspected,reproductiveanddevelopmentaltoxicityassociatedwith formaldehydeexposureremainsinconclusive.
1.1. Previousreviewsofformaldehydereproductiveand developmentaltoxicity
Reproductive toxicity broadly refers to the occurrence of biologicallyadverseeffectsonthereproductivesystemthatmay result fromchemical exposure to environmental agents and is characterizedbyalterations tothefemaleormalereproductive organs, related endocrine system, or pregnancy outcomes [6].
Likewise,developmentaltoxicity(alsoknownasteratogenicity)is theoccurrenceofadverseeffectsonthedevelopingorganismthat may resultfromchemical exposure priortoconception, during prenatalorpostnataldevelopment,andmaybedetectedatany pointinthelifespanofanorganism.Majormanifestationsinclude deathofthedevelopingorganism,structuralabnormality,altered growth,andfunctionaldeficiency[6].
Earlyreviews,ofteratogenicityofformaldehydeinanimals[7], andteratogenicityandreproductivetoxicityofformaldehydeinboth animalsandhumans[8,9],concludedthattheevidencewaslimited andwasfromasmallnumberofstudies,which,inthecaseofthe humanstudieswereoftenofpoorquality,lackingaccurateexposure informationandstatisticalpower.Onelimitationidentifiedinmany early animal studieswas thatthe effects offormaldehyde were assessedindirectlythroughitsmetabolismfromhexamethylenetet- ramine,whichisconditional.Inits2006monographonformalde- hyde, IARC found that existing studies of formaldehyde’s reproductiveanddevelopmentaleffectsinbothhumansandanimals wereinconclusive,notingthatmostoftheepidemiologicalstudies reviewedwerenotdesignedtospecificallyevaluateformaldehyde, andthatmoreexposure-specificfollow-upstudieswererequired[1].
The US EPA, in a draft document reviewing formaldehyde inhalation toxicity in animals and humans, suggests that the developingorganismandthereproductivesystemaretargetsfor toxicityfollowingformaldehydeexposurebyinhalation,although thesefindingsaresubjecttorevisionaspartofEPA’songoingreview process [5].Theanimalstudiesexamineddemonstrated abroad rangeofadverseoutcomesfollowingexposure,whilehighlighting theinadequacytoassesstheseoutcomes.Sincesimilaroutcomes
werealsoobservedinhumanstudies,theoveralldatasupportedthe humanrelevanceofreproductiveanddevelopmentaltoxicity.This reviewalsodiscusseddatagapsinthecurrentliterature,suchasa lackofassessmentofpotentialreproductiveeffectsinhumanmales [5].Inthemostrecentreviewofformaldehydereproductivetoxicity in2001,Collinsetal.concludedthatthe reproductiveimpactof formaldehydeinhumanswas unlikelyat occupational exposure levels,despitefindingevidence ofincreasedriskofspontaneous abortion(SAB)inameta-analysisof8humanpopulationstudiesof formaldehyde-exposedworkerswhichreportedsufficientdata[10].
Further, it was concluded that there was little evidence of reproductiveor developmentaltoxicityat levelsof occupational formaldehydeexposureintheanimalstudies,althoughroutesof exposureotherthaninhalationweredisregarded.
Weconductedthepresent reviewtoprovidea comprehen- sive, updated assessment of reproductive and developmental toxicity, particularly adverse pregnancy outcomes, associated withformaldehydeexposure.
1.2. Systematicapproachofthecurrentreview
Inthisreview,mostofthepublishedstudiesinexposedhumans and experimental animals are reviewed; findings from a new meta-analysis of human epidemiology studies are reported;
additionalandrelevantevidencefrominvivoandexvivoanimal studies are examined; and potential mechanisms of action of formaldehyde-inducedreproductive and developmentaltoxicity arediscussed.
Electronic searches were performed on PubMed using key- words including: formaldehyde, formalin, formol, reproductive toxicity, developmental toxicity, embryotoxicity, teratogenicity,
andpregnancyoutcomes.Searchesincludedcase-control,nested case-control,cross-sectional,andcohortstudiesinhumans,aswell asstudiesconductedviaanyrouteofformaldehydeexposureat anydosageonanyexperimentalanimalspecies.Weadditionally cross-referenced other formaldehyde reviews and books to generate a more complete list of literature. Collaborators from ourprevious reviewofformaldehyde inChina[2] wereableto obtain additional papers from the China National Knowledge Infrastructure, which contains 7426 Chinese-language journals from 1915 to the present, an otherwise inaccessible source of information. We systematically excluded studies published in languagesotherthanEnglishandChinese,andstudiespre-dating 1980duetodifficultiesinacquisition,unlessthestudiespresented human data, for which there were already limited resources.
Studiesforwhichfulltextpublicationswereunobtainablewere alsoexcluded.Weconcentratedonstudieswithclearanddirect formaldehyde exposure, and did not include those in which formaldehydewasabyproductofexposuretoanotheragent(e.g.
formaldehyde-releasingprodrugsandcosmetics, orhexamethy- lenetetramine and aspartame). The study selection process is detailedinFig.1.
2. Humanpopulationstudies
Weidentified18humanstudiesreportingonthereproductive effectsofformaldehyde-exposedpopulations.Inallbut2studies, women were chronically and/or occupationally exposed to formaldehydeeitherbeforeorafterconception,andtheoutcomes examinedincludedmenstrualabnormalities,infertility,spontane- ous abortions, stillbirths, congenital malformations, premature birth, and birthweight. Theremaining 2 studiesexamined the Removed if full text not available
64 References for Detailed Evaluation
Removed irrelevant studies based on abstract & full text
Removed non-English publications, except Chinese or
human studies
Removed studies pre-dating 1980 unless human study
Removed repeated citations
18 Human Studies 46 Animal Studies
PubMed Keyword Search formaldehyde, formalin, formol, reproductive toxicity,
developmental toxicity, embryotoxicity, teratogenicity,
pregnancy outcomes
Cross Referencing Other Review Papers Staples 1983; Feinman 1988;
WHO 1989; IARC 2006;
Collins 2001; NTP 2010;
US EPA 2010
450 Potential References
Chinese Studies Most recently peer-reviewed
formaldehyde and reproductive toxicity studies
published in Chinese
Fig.1.Flowdiagramofstudyselectionprocess.Thisfiguredepictsthelogicofthestudyselectionprocess,theresultsofwhichareincludedforreviewinthispaper.
A.Duongetal./MutationResearch728(2011)118–138 120
reproductiveeffectsofpaternalexposuretoformaldehyde,with onestudyanalyzingspermmorphologyinexposedmaleworkers [11], and the other investigating risk of spontaneous abortion resultingfrompaternalexposure[12].Findingsfromthe18human studies are summarized in Table 1. Studies are categorized by outcome and listed chronologically for each outcome. Because severalstudiesreportmultiplereproductiveand developmental outcomes,theyarecitedseveraltimesinTable1andthroughout thetext.
2.1. Reproductivetoxicityinhumans
Alteredincidences ofpregnancies,abnormalmenstruationor abnormal sperm may each serve as a potential indicator of reproductivetoxicityinhumans.Ina1975Russiancross-sectional study,menstrualdisorderswerereported2.5timesmoreoftenin womenoccupationallyexposedtoformaldehydethanincontrols [13]. A later Danish cross-sectional study examined menstrual irregularitiesin7mobiledaycarecentersinwhichaverageindoor formaldehydeconcentrationsmeasured0.43mg/m3or0.35parts permillion(ppm)duetotheuseofurea formaldehydeintheir construction[14].Menstrualirregularitieswereself-reportedin 30–40%ofthefemaleexposedworkers,comparedtononeinthe matched unexposed control group. The exposed group also experiencedgreatervaginalirritationandpainduringmicturition (urination).
AFinnishcohortstudyinvestigatedtheeffectofformaldehyde onfemalefertilityasmeasuredbyfecundabilitydensityratio(FDR) [15].AnFDRsignificantlybelow1.0meansdelayedconception,an indicatorofreducedfertility.Exposuretohighlevelsofformalde- hyde(mean=0.33ppm)wassignificantlyassociatedwithdelayed conception; the adjusted FDR was 0.64 with 95% confidence interval(CI)0.43–0.92forthehighexposedgroupcomparedtothe control unexposed group. This cohort study also found an increasedrisk of endometriosis, withanodds ratio (OR)of 4.5 and 95% CI of 1.0–20.0, further suggesting that formaldehyde exposure may have an adverse effect on female reproductive affects.
In contrasttothestudies infemales, in theonly studythat examined malereproductive effects, a Finnishcohort study,no adverseeffects on spermproduction, suchassperm countand sperm morphology, were found to be statistically different between exposed and unexposed groups [11]. However, as acknowledgedbytheauthors,giventhesmallsizeoftheexposure groups(n=11ineachgroup)andthelargestandarderrors(SE)in thecontrolgroup,thestudyhadverylowstatisticalpower.
2.2. Developmentaltoxicityinhumans
Developmentaltoxicitydescribestheabilityofasubstanceto causeadverseeffectsinthedevelopingorganism,withmanifesta- tionsincludingspontaneousabortion,stillbornbirths,congenital malformations and other structural abnormalities, low birth weightandprematurebirths(Table1).
2.2.1. Spontaneousabortion
Spontaneousabortion,alsoknownasmiscarriage,isdefinedas apregnancythattypicallyendsnaturally(notinduced)duringthe first7–28weeksofgestation,andoccursatarateof15–20%inthe UnitedStatesandatlowerratesinmostdevelopedcountries[16].
The majority of studies on SAB associated with formaldehyde exposureexaminedtheeffectofmaternalexposure,withonlya singlestudyexaminingpaternalexposure.
Toourknowledge,theearlieststudyofdevelopmentaltoxicity inhumanswasconductedin 1975,in Russia,onoccupationally
exposedfemalefactoryworkers[13].Nodifferenceintherateof Table1 Summaryofreproductiveanddevelopmentaltoxicityinhumanstudies. aaOutcome&studyStudytypeStudypopulationExposuresourceLocationFormaldehydeOutcome(95%CI) bconcentration Fertility 3Shumilina(1975)[13]Cross-sectional130highexposed,316lowexposed,200FemalefactoryworkersRussia1.5–4.5mg/mMenstrualdisordersinFAexposed2.5higherthan unexposedintheunexposed 3Olsen&Dossing(1982)[14]Cross-sectional66exposed,26unexposedFemaledaycareworkersDenmark0.43mg/mMenstrualirregularitiesinabout30–40%ofFA exposedand0%ofunexposed;higherrateof vaginalirritation&painduringmicturition Taskinenetal.(1999)[15]Cohort119lowexposed,77moderate,39high,FemalewoodworkersFinland0.01–1.00ppmReducedfertility,FDR=0.64(95%CI0.43–0.92), 367unexposedp0.02athighFAexposure(mean0.33ppm) Wardetal.(1984)[11]Cross-sectional11exposed,11unexposedMaleautopsyworkersTexas0.1–5.8ppmNostatisticallysignificantdifferenceinspermmorphology orcountbetweenFAexposedandunexposed SpontaneousAbortion 3Shumilina(1975)[13]Cross-sectional130highexposed,316lowexposed,FemalefactoryworkersRussia0.05–4.5mg/mNodifferencebetweenFAexposedandunexposed 200unexposedin‘‘prematurebirthsorabortions’’ Hemminkietal.(1982)[17]Cohort1100unexposedpregnancies,50FemalehospitalworkersFinlandN/AAdjustedratesof8.4%SABinFAexposedand formaldehydeexposedpregnancies.8.3%inunexposed Axelssonetal.(1984)[19]Cohort1160totalpregnanciesamong745FemalelabworkersSwedenN/A30%SABrateinFAexposed,11.5%inunexposed females,10exposed cHemminkietal.(1985)[18]Case-control164cases,464controlsFemalehospitalnursesFinlandN/ACOR:0.7(0.28–1.73)ofSABforFAexposure;6 exposedincasesand24exposedincontrols Seitz&Baron(1990)[20]Cohort365pregnanciesFemalefabricworkersKentucky0.14–0.79ppmMiscarriagerate:fabricworkersatRockcastle,14%;working (Rockcastlefacility)elsewhere,13%;notworkingoutsidehome,5%.
Table 1 (Continued)
Outcome & studya Studyatype Study population Exposure source Location Formaldehyde
concentrationb
Outcome (95% CI)
Stucker et al. (1990)[21] Cohort 139 exposed (cytostatic drug) pregnancies, 357 unexposed; 113 FA-exposed or exposure unknown
Female hospital nurses France N/A FA did not modify the risk of SAB associated with
exposure to cytostatic drugs
John et al. (1994)[22] Cohort 244 formaldehyde exposed and 132
unexposed
Cosmetologists US N/A AOR: 2.1 (1.0–4.3) of SAB for FA exposure
Saurel-Cubizolles et al.
(1994)[23]
Cohort 724 pregnancies: 316 formol exposed, 408 unexposed
Female hospital nurses France N/A 11.1% SAB in formaldehyde exposed, 6.9% in unexposed
Taskinen et al. (1994)[24] Case-control 206 cases, 329 controls Female lab workers Finland 0.01–7 ppm OR: 3.5 (1.1–11.2) of SAB for FA exposure
Taskinen et al. (1999)[15] Cohort 52 cases with same work place from preconception through year of SAB, in high, medium and low exposure categories
Female wood workers Finland 0.01–1.00 ppm OR 3.2 (1.2–8.3) of SAB for high FA exposure
Lindbohm et al. (1991)[12] Cohort 596 formaldehyde exposed pregnancies, 54 SAB cases
Paternal occupational exposure
Finland ‘‘Moderate/High’’ OR: 1.0 (0.8–1.4) of SAB for FA exposure Lindbohm et al. (1991)[12] Cohort 1212 formaldehyde exposed pregnancies,
110 SAB cases
Paternal occupational exposure
Finland ‘‘Low’’ OR 1.1 (0.9–1.4) of SAB for FA exposure
Congenital Malformations
Ericson (1984)[26] Nested
case-control
26 cases, 50 controls Female lab workers Sweden N/A OR = 0.96 (0.22–4.18) of SAB for FA exposure;
3 exposed in cases and 6 in controls
Hemminki et al. (1985)[18] Case-control 34 cases, 95 controls Female hospital nurses Finland N/A COR: 1.74 (0.40–7.60)dof CM in FA exposed;
8.8% cases FA exposed, 5.5% controls Saurel-Cubizolles et al.
(1994)[23]
Cohort 641 pregnancies: 271 exposed, 370 unexposed
Female hospital nurses France N/A 5.2% CM in FA exposed, 2.2% in unexposed
Dulskiene & Grazuleviciene (2005)[27]
Case-control 184 cases, 479 controls Ambient exposure Lithuania >2.4mg/m3 OR = 1.24 (0.81–2.07) of heart CM for FA exposure
Zhu et al. (2006)[28] Cohort 983 pregnancies: 218 high exposure, 364 medium exposure, 401 low exposure
Female lab workers Denmark N/A AOR: 1.5 (0.8–2.9) of CM for high FA exposure
Low Birth Weight
Shumilina (1975)[13] Cross-sectional 130 high exposed, 316 low exposed, 200 unexposed
Female factory workers Russia 0.05–4.5 mg/m3 No cases in high-exposed group, 2 cases in low-exposed group, 2 cases in control group Grazuleviciene et al.
(1998)[29]
Cross-sectional 4343 births, 244 cases Ambient exposure Lithuania <1.94 to>4.67mg/m3 Adjusted RR 1.37 (0.90–2.09) of low birth weight for FA exposures>3.5 ug/m3compared to<3.5 ug/m3 Maroziene & Grazuleviciene
(2002)[30]
Cross-sectional 3988 births, 140 cases Ambient exposure Lithuania Mean 3.14mg/m3and
tertiles<2.00, 2.01–3.9
and>3.9mg/m3e
OR adjusted for gestational age 2.09 (1.03–4.26) of low birth weight for>3.9 ug/m3FA Zhu et al. (2006)[28] Cohort 983 pregnancies: 218 high exposure,
364 medium exposure, 401 low exposure
Female lab workers Denmark N/A AOR: 1.2 (0.6–2.2) of low birth weight at high
FA exposure index Preterm birth
Shumilina et al. (1975)[13] Cross-sectional 130 high exposure, 316 low, 200 unexposed
Female factory workers Russia 1.5–4.5 mg/m3 No difference between exposed and unexposed in ‘‘premature births or abortions’’
Maroziene & Grazuleviciene (2002)[30]
Cross-sectional 3988 births, 203 cases Ambient exposure Lithuania Mean 3.14mg/m3and
tertiles<2.00, 2.01–3.9
and>3.9mg/m3e
AOR: 1.37 (0.91–2.05) of preterm birth for FA exposure
Zhu et al. (2006)[28] Cohort 983 pregnancies: 218 high exposure, 364 medium exposure, 401 low exposure
Female lab workers Denmark N/A AOR: 0.7 (0.3–1.7) of preterm birth at high FA
exposure index Other developmental toxicity
Seitz & Baron (1990)[20] Cohort 365 pregnancies Female fabric workers Kentucky 0.14–0.79 ppm RR: 6.9 (3.6–13.1) of stillbirth, premature birth,
birth defect for FA exposure
Abbreviations: AOR, adjusted odds ratio; CM, congenital malformation; COR, crude odds ratio; CRR, crude relative risk; FA, formaldehyde; FDR, fecundability density ratio; OR, odds ratio; N/A, not available; ppm, parts per million; RR, relative risk; SAB, spontaneous abortion; TWA, time weighted average.
a Categorized by pregnancy outcome, then listed chronologically by publication year.
b Values are as reported, not converted, 1 ppm = 1.23 mg/m3.
cThe reported crude OR is 0.6, but a crude OR of 0.70 is calculated based the data in theirTable 2.
d The reported crude OR is 1.8, but a crude OR of 1.74 is calculated based on the data in their Table 6.
e Obtained from additionalTable 1in[30].
A.Duongetal./MutationResearch728(2011)118–138122
abortionbetweenexposedandunexposedworkerswasfoundin thiscross-sectionalstudy.
A1982retrospectivecohortstudyofhospitalstaffmembersin Finland found that formaldehyde exposure at concentrations found in Finnish hospital sterilization units (typically 0.03–
3.5ppm; not measured in this study specifically) was not associatedwithanincreasein SABbasedonanalysisofasmall numberofformaldehyde-exposedwomen(n=50exposedpreg- nancies)and1100unexposedpregnancies[17].Theadjustedrates were8.3%forunexposedpregnanciesand8.4%forformaldehyde- exposedpregnancies.Thesameresearchgroupconductedamore in-depth case-control study of Finnish hospital nurses and reconfirmed that there was no relation between formaldehyde exposureandSAB[18].
Axelsson et al. (1984) interviewed 745 Swedish female universitylaboratoryworkerswhohadatotalof1160pregnancies [19].Inthiscohortstudy,therewasaslightlyelevatedrelativerisk (RR) ofmiscarriageratein women exposedtoorganic solvents duringtheirfirsttrimester(RR1.31,95%CI0.89–1.91).Amongthe 10 women specifically exposed to formaldehyde,there were 5 normalbirths,2inducedabortionsand3miscarriages,thus,the miscarriageratewas30%.Thecorrespondingmiscarriageratefor womenwhodidnotconductlaboratoryworkwhilepregnantwas 11.5%.Axelssonetal.foundthatexposuretoformaldehydeduring pregnancyshowedthehighestmiscarriageratecomparedtoother volatileorganiccompoundsbutthenumberofcaseswastoosmall toconcludeadefinitivecausalrelationship.
AstudybytheNationalInstituteforOccupationalSafetyand Health(NIOSH)examinedtheoutcomesof365pregnanciesina cohortof407femaletextileworkersinafacilitythatfabricated men’sworkpantsinKentuckyandfoundthatthemiscarriagerate in those who worked in the facility while pregnant(14%) was similar to the rate of those who worked elsewhere during pregnancy(13%) [20]. Althoughtheseratesweresimilar tothe rateofmiscarriageinthegeneralpopulation (10–25%),theSAB rateamongthosetextileworkerswhodidnotworkoutsideofthe homeduringpregnancywasonly5%.
InaFrenchcohortstudyexaminingSABamonghospitalnurses handling neoplastic drugs, formaldehyde was assessed as a confounding exposure [21]. Data were collected by interview fromMay1985toMay1986inthreeFrenchhospitalsandinalarge center for cancer treatment. Of 139 pregnancies in nurses occupationallyexposedtocytostaticagents(whichsuppresscell growth and multiplication), the frequency of SAB was 25.9%
comparedtoonly15.1%inthe357unexposedpregnancies(RR1.7, 95% CI1.2–2.5)[21].When thepregnancies identifiedas being positiveorunknownforpreviousformaldehydeexposure(n=113) were excluded, the results concerning cytostatics were not modified.Thesedataindicatethatformaldehydedoesnotinteract withcytostaticdrugstocauseSABs,buttheeffectofformaldehyde alonewasnotanalyzed.
AnationwidedatabaseofmedicallydiagnosedSABwasusedto evaluatetheeffectsofpaternaloccupationandexposureonSAB riskinFinland.Inacohortof596pregnancies,anadjustedORof 1.0, 95% CI 0.8–1.4 of SAB wasfoundfor paternal exposure to moderate or high formaldehyde concentrations [12], and an adjustedORof1.1(95%CI0.9–1.4)forlowformaldehydeexposure, showingnooverallexcessofSABinwomenwhosehusbandswere exposedtoformaldehyde.Theauthorshypothesizedthatifthere hadbeenanymale-mediatedeffectsonpregnancyoutcome,the onlypossibledamagewouldbeviageneticdamagetomalegerm cellsorbysecondarymaternalexposure.Sinceindividualexposure couldnotbeassesseddirectly,anyconclusionsaboutthisstudyare purelysuggestive.Theauthorsrecommendthatthefindingsofthis study‘‘needtobeconfirmedbystudiesinwhichindividualexposures canbeassesseddirectly’’[12].
Toexaminetherelationbetweenadversepregnancyoutcomes incosmetologistswhoareoftenexposedtoavarietyofchemicals, including formaldehyde-based disinfectants, a cohort of female cosmetologists from North Carolina were surveyed, and it was foundthatfull-timecosmetologistswhousedformaldehyde-based disinfectants had a 2.1-fold(95%CI 1.0–4.3)higher riskof SAB than those who did not useformaldehyde-based disinfectants, when adjusted for other chemical exposures and maternal characteristics[22].
InacohortstudyofFrenchhospitalworkers,itwasfoundthat therateofSABwassignificantlyhigheramongpregnanciesduring whichwomenworkedinanoperatingroomandwereexposedto formol (10%formaldehydesolution),ionizingradiation oranes- thetics. Of the724total pregnancies,11.1% of thepregnancies exposedtoformolresultedinSAB,comparedtoonly6.9%inthe unexposed group(p<0.05)[23]. However, asdiscussed bythe authors,exclusionof theeffects ofexposuretochemicalsother thanformaldehydeintheoperatingroomswasnotpossible.
InFinland,agroupofscientistsidentifiedSABcasesinanation- widecohortofwomenworkinginlaboratorysettings.Comparedto unexposedwomen,womenwhoworkedinlaboratoriesandwere chronicallyexposed(3–5days/week)toformalin,a37%formalde- hydesolution,showedanincreasedriskforSAB(OR3.5,95%CI 1.1–11.2) [24]. The same group of researchers also found an increasedriskforSAB(OR3.2,95%1.2–8.3)amongfemalewood workerswhowerechronicallyexposedtoformaldehydeathigh levelsinacase-controlstudy[15].
2.2.2. Congenitalanomalies
Congenital anomalies, or birth defects, are characterized by physical, metabolic or anatomic deviations from the normal patternofdevelopmentthatareapparentatbirthandaffecthowa babywilllook,function,orboth.Theyrangefrommildtofatal,and affectabout3%ofallbabiesbornintheUS[25].Veryfewstudies haveexaminedcongenitalanomaliesandformaldehydeexposure.
Ericsonetal.reportedahigherthanexpectednumberofinfants whodiedneonatallyand/orhadcongenitalmalformationsamong birthsinlaboratoryworkerscomparedtoallotherbirthsbasedon datafromthe1975Swedishcensusandthe1976MedicalBirth Register[26].Inthisstudy,nodatawasavailableonformaldehyde exposure and no specific type of laboratory work could be identified to be more common among those with abnormal pregnancyoutcomesthan thenormalcontrols. Ina smallcase- controlstudynestedwithinthislargerstudy,noassociationwas seenwithformaldehydeexposureinthefemalelaboratoryworker mothersof26infantsingletonswhohaddiedorhadmalforma- tionscomparedwith50randomlyselected,age-matchedcontrols.
In thisnestedcase-controlstudy,qualitativeexposuredatawas obtainedbyquestionnaire(i.e.subjectswereaskedtolistharmful substancestowhichtheywereexposed)andwasthereforesubject topossiblerecallbias.
In the aforementioned Hemminki et al. (1985) case-control study, while SAB risk was not increased, risk for congenital malformations was increased among children born to female hospitalnurseswithformaldehydeexposure.Amongthe34cases of malformed children three (8.8%) were born to women who reported formaldehyde exposure during their first trimester.
Among controls, 5 of 95 (5.3%) working women reported formaldehyde exposure [18]. The authors noted that because theirstudyhadlowstatisticalpower,onlyverypotenteffectscould havebeenidentified.
Inadditiontospontaneousabortion,theFrenchcohortstudy conductedbySaurel-Cubizollesalsoinvestigatedbirthdefectsin babies born to female hospital nurses with and without formaldehyde exposure. Of 641 total pregnancies,there wasa greaterfrequencyofbirthdefectsinthepregnanciesexposedto
formol(5.2%)thanthosewhowereunexposed(2.2%)[23].Itwas also found that formol caused the highest frequency of birth defectsamong alltheexposure agents,suchasanestheticsand ionizingradiation, investigated in this study. In a newer study conductedinLithuania,itwasfoundthatresidenceinanareawith ambientformaldehyde>2.4
m
g/m3wasassociatedwithincreased congenitalheartmalformationby24%(OR1.24,95%CI0.81–2.07) [27].ADanishcohortstudyfoundthattheadjustedhazardratioof highformaldehydeexposureand‘major’malformationswas1.5 (95%CI0.8–2.9)[28].Taskinen et al. alsofoundno increase in ORs forcongenital malformationsdue tomaternalexposure tosolventsin general throughworkinalaboratory,butdidnotexamineformaldehyde exposurespecifically[24].
2.2.3. Lowbirthweight
A population-based non-occupational study in Lithuania compared low birth weight (<2500g) rates among women residinginareaswithhighorlowconcentrationsofformaldehyde inambientair[29]. Inthis cross-sectionalstudy,thecrude risk ratiooflowbirthweightbabiesamongwomenresidinginhigh formaldehydeexposureareas(>4.67
m
g/m3)was1.68with95%CI 1.24–2.27comparedwithwomenresidinginlowexposureareas (>4.67m
g/m3).Onceadjustedforpotentialconfounders,theOR was1.37(95%CI0.90–2.09)forexposures>3.5m
g/m3comparedto <3.5
m
g/m3. Increasing levels of formaldehyde exposureresultedin increasedincidence of low birth weight, with48.3 per1000,49.5and81.1,inlow,moderate(>3.48
m
g/m3)andhigh- exposureareas,respectively.Thesameresearchgroupconducteda follow-upstudyofallnewbornsbornin1998inthecityofKaunas, Lithuania[30].Residentialexposurelevelsweremonitoredat12 municipalmonitoring sites,oneineach residentialdistrict,and logisticregressionwasusedtoestimatetheeffectofpollutantson reproductive outcomes. The adjusted OR for low birth weight (<2500g)atthehighestambientformaldehydelevelwasORof 2.09(95%CI,1.03–4.26).ThisORwasadjustedforparity,maternal age,marital status,education,season, smoking,and gestational age.TheunadjustedORwaslowerandnotstatisticallysignificant (1.39;95%CI,0.91–2.12).InaDanishstudy,theadjustedORforlow birthweightfor motherswhowerelaboratorytechnicianswith highformaldehydeexposurewas1.2(95%CI0.6–2.2)[28].2.2.4. Prematurebirth
Shumilinaetal.foundthattherateofprematurewaterbreaking was37.232.41%inRussianfemalefactoryworkersoccupationally exposedtoformaldehyde,comparedto23.631.23%intheunexposed group,butnoinformationonsignificancewasprovided.Additionally, thethreatofintra-uterinefetalasphyxiation,aconditioninwhichthere isanextremedecreaseinoxygensupplytofetuses,wasmorethan2 timeshigherintheexposedgroupthaninthecontrolthoughnoactual dataonoxygenlevelswerereported[13].Prematurebirthratesdidnot differbetweenexposedandunexposedgroups.
MarozieneandGrazulevicienestudiedtheeffectsof ambient formaldehydeandprematurebirthina cross-sectionalstudyof 3988birthsandfoundthatathighambientformaldehydelevels, theriskofprematurebirthwas1.37(95%CI0.91–2.05)[30].
Themostrecentepidemiologystudyonformaldehydeexposure andpregnancyoutcomesidentifiedandsurveyedacohortoffemale laboratoryworkers[28].Areducedriskofpre-termbirth,OR0.7 (95%CI0.3–1.7),wasfoundforthosewhoreportedlaboratorywork involvingfrequentand/orhighformaldehydeexposure.
2.2.5. NIOSHanalysisofcombinedadversebirthoutcomes
In1987,NIOSHwasrequestedtoconductahealthevaluationat Rockcastle Manufacturing,a textileplant that fabricatedmen’s workpants in Kentucky [20]. Employees were complaining of
headaches,nausea,vomiting,fainting, andadversereproductive effectsatthefacility.Formaldehydeairsamplingresultsranged from0.32to0.70ppmin13airareasamplesobtainedthroughout the plant, lower than the current US Occupational Safety and Health Administration (OSHA) occupational exposure limit of 0.75ppm [31]. Additionally, fabric samples that the company produced released 163–1430 micrograms of formaldehyde per every gramof fabric (
m
g/g). Pastand present employees were surveyed for health information and reproductive health data (including miscarriage, birth defects, premature births and stillbirths)wereassessed.Theresponserateswere98%forcurrent employees and only 18% for past employees. A total of 365 pregnanciesweredividedinto3categories:(1)pregnanciesthat occurred while the woman was employed at Rockcastle; (2) pregnanciesthatoccurredwhilethewomanworkedelsewhere;(3) and pregnancies that occurred while the woman was not workingoutsidethehome.Theratesofbirthdefects,stillbirthsand prematurebirthscombinedamongworkersincategories(1),(2) and(3)were42%,5%and6%,respectively.TheRRofhavinganyof theseadversepregnancyoutcomesincategory(1) comparedto thosepregnanciesincategories(2)and(3)combinedwas6.9(95%
CI,3.6–13.1,p<0.001)[20].Theratesofmiscarriageingroups1,2, and3were14%,13%,and5%,respectively.Theauthorsnotedthat therateofmiscarriageingroup3(thosenotworkingoutsidethe home) were2–3 timesbelow national averages. Thislow rate, combinedwiththelackofdetailsregardingthemethodsused,and theuseofexposedwomenastheir ownunexposedcomparison group,makestheresultsofthisstudydifficulttointerpret.
2.3. Limitationsofthehumanstudies
Among the 18 human studies identified, there were more developmental studies than reproductive toxicity studies, likely because developmental toxicity has greater and more obvious physicalmanifestations,whereasreproductivetoxicityeffectsare more difficult todetect and determine.The studies suffer from limiteddesignandscope,andthusdonotconclusivelydetermine whetherformaldehydeexposurecauseshumanreproductiveand developmentaltoxicities.Manyoftheolderstudiesreliedonself- reporteddata,andmaysufferfromreporting,recall,andselection biases.Astheywerepredominantlyretrospectiveepidemiological studies,fewprovidedlevelsofformaldehydeexposurebecausethey were not specifically designedtoevaluate this. In addition, the resultsobtainedmayhavebeenconfoundedbyotherco-exposures.
Noneofthestudiesofferaplausiblebiologicalmechanismbywhich reproductiveanddevelopmentaltoxicitycouldoccur.
There isanoverwhelminglylargerportion ofhumanstudies examiningreproductiveoutcomesassociated withfemaleexpo- suresand thusadearthofstudiesassessingpotentialeffects of formaldehyde exposure in males. More human studies of reproductiveeffectsresultingfromexposureinmalesareneeded, inordertounderstandmaleinitiatedmechanisms.
In summary, despite study design limitations, this brief evaluationoftheprevioushumanstudiesprovidesatleastsome evidence that formaldehyde exposure may be associated with reproductiveanddevelopmentaltoxicity,whetherimpactingone or multiple reproductive outcomes. To further evaluate the association between formaldehyde association and these out- comes,weconductedanupdatedmeta-analysis.
3. Updatedmeta-analysis
3.1. PreviousresultsfromCollinsetal.
To date, only one other meta-analysis has examined the relationship between formaldehyde exposure and adverse A.Duongetal./MutationResearch728(2011)118–138
124
pregnancy outcome [10]. This meta-analysis, by Collins et al., included 8 epidemiology studies with data on occupational formaldehydeexposureandspontaneousabortion.Collinsetal.
reportedasummaryRRof1.4(95%CI0.9–2.1),whichtheauthors stated‘‘showedsomeevidenceofincreasedrisk’’.However,the authors also identified evidence of publication bias (i.e. the tendency of researchers and journals not to publish smaller studieswithnegativeornullresults).Evidenceofpublicationbias wasseeninthefunnelplotofthestudy’seffectsizes(e.g.theOR) versustheirsamplesizes,in Beggs test,andintheirsubgroup analysesbased onstudysize (that is, thesummary RR in the studieswith40ormoreexpectedcaseswas0.7(95%CI0.5–1.0) althoughtheanalysisofthesmallerstudiesonlyincludedtwo studies.Inaddition,partofthereasonthesetestsmayhaveshown someindicationofpublicationbiaswastheinclusionofthelarge negativestudyofpaternalexposuresbyLindbohmetal.[12].The authorsalsoraised thepossibilityofrecallbiasandperformed subgroupanalysesbasedonwhetherformaldehydeexposurewas determinedusingself-reporteddata.ThesummaryRRforthe5 studiesusingself-reportedexposuredatawas2.0(95%CI,1.4–
2.8).Incontrast,thesummaryRRforstudiesusingothermethods ofexposure assessmentwaslower (summaryRR=0.7;95%CI, 0.5–1.0), although there were onlytwo studies in this group.
Basedonthepossibilityofrecallandpublicationbias,theauthors interpreted their results as negative and concluded that occupationalexposuretoformaldehydedidnotincreaseriskof spontaneous abortion.The results werefurther interpreted to conclude that formaldehyde is unlikely to cause any adverse pregnancyoutcomesatoccupationalexposurelevels.
3.2. Uniqueapproachinthepresentanalysis
The present meta-analysis differs from the previous Collins etal.meta-analysisinseveralways.First,weperformedseparate meta-analyses for SAB and for all developmental outcomes combined,whereasCollinsetal.onlypresentedresultsforSAB.
Wegroupedtheseoutcomestogetherforanalysistoincreasethe powertodetectdevelopmentaloutcomesgenerallyandbecause theymaypotentiallyresult fromeffects ofexposure on similar targetsorpathwaysduringthecriticalpreconceptionwindow,e.g.
genotoxicdamagetogermcells.Second,Collinsetal.combined studiesofmaternalandpaternalexposureintheirmainanalyses, while our main analyses only included studies of maternal formaldehydeexposures.Third,weincludedthestudyofoperating roomnursesbySaurel-Cubizollesetal.(1994),whichidentifieda statisticallysignificantincreaseofSABinformol-exposednurses (11.1% vs. 6.9%, p<0.005; calculated COR=1.68, 95% CI 1.01–
2.82).This paper,however,wasnotmentioned inCollins etal.
Fourth,whenrelativerisksweregivenforseveraldifferentlevelsof exposure(e.g.low,medium,andhigh),weusedtherelativeriskfor thehighestexposurelevel.Incontrast,Collinsetal.usedtheRRfor all exposure levels combined. If true associations exist, higher
exposuregroupsaregenerallyassociatedwithgreaterstatistical power and less likelihood of important confounding [32].
Importantly,thisdifferenceonlyappliedtoonestudy:Taskinen et al., 1999 [15]. Fifth, we excluded the study of SAB and antineoplastic drugs by Stucker et al., 1990 [21], which was includedinCollinsmeta-analysis,becauseofthelargenumberof peopleforwhomformaldehydeexposurewasunknown(50people had known formaldehyde exposure, whereasfor 63 people the formaldehydeexposurewasunknown).Finally,therewereseveral mostlyminordifferencesinthemethodsusedtocalculatecrude ORsandconfidenceintervalswhenonlyraw22tabledatawere provided.
3.3. Meta-analysismethods
From the18 humanstudies identified, certain studieswere excludedfromthemeta-analysisifRRsorestimatesofvariance werenotprovidedorcouldnotbeestimated[11,13,14],orifthe studydidnotincludeanindependentgroupofunexposedcontrols [20],ordidnotprovideformaldehydeexposuresforthemajorityof exposedsubjects[21].Theexcludedstudiesandreasonsfortheir exclusionaresummarizedinTable2.Asdiscussed,ifdifferentRRs were presentedfor different levelsof exposure, theRR for the highest exposure category was used in the meta-analysis [15,29,30].
Meta-analysesweredonefortwooutcomescategories:SABand all reproductiveand developmentaloutcomescombined,which includeSAB,birthdefectsormalformations,andlowbirthweight.
SABwastheonlyindividualoutcomewithanadequatenumberof studies(n=8;Table1)toperformameta-analysis.Severalstudies provideddataformorethanoneoutcome.Inordertohelpassure independenceacrossstudies,inthemeta-analysisofalloutcomes combined,arelativeriskforasingleoutcomewasselectedfrom eachstudyinthefollowingorder:SAB,birthdefects/malformation, andlowbirthweight.SABandbirthdefects/malformationswere chosenfirstandsecondbecausethesewerethefirstandsecond mostcommonindividualoutcomesassessed.Allbutoneselected study [12] assessed formaldehyde exposure in the mother.
Separate analyses were done with and without this study to assessitsimpactonoverallresults.
MicrosoftExcel2008andSTATAversion11 (College Station, Texas)wereusedforallcalculations.SummaryRRestimateswere calculatedusingboththefixedeffectsinversevarianceweighting methodand therandom effects method[33,34].Heterogeneity wasevaluatedusingthegeneralvariance-basedmethod[35].If heterogeneityispresent,therandomeffectsmodelincorporates between-studyvariationintothesummaryvarianceestimateand confidence intervals. Some authors have suggested that the randomeffectsmodelmaybemoreconservative[35].However, unlike the fixed effects model, where weights are directly proportionaltostudyprecision,therandomeffectsmodelweighs studiesbasedonahighlycomplexandnon-intuitivemixofstudy
Table2
Studiesnotincludedinthemeta-analysisandrespectiveexclusioncriteria.
Study Outcome Result(RRandCIorother) Reasonnotused
Shumilina(1975)[13] Lowbirthweight(<2500g) Nocasesinthehighexposedgroup, 2casesintheunexposedgroup.
NoRR,poorlydescribedmethods
Shumilina(1975)[13] Menstrualirregularities 47.5%vs.18.6% Poorlydescribedmethods
Shumilina(1975)[13] PrematurebirthsandSAB ‘‘nodifferences...’’ NoRR,poorlydescribedmethods Olsen&Dossing(1982)[14] Menstrualirregularities Menstrualirregularitiesinabout30–40%
ofexposedand0%ofunexposed
NoRR Wardetal.(1984)[11] Spermcountandmorphology ‘‘nostatisticallysignificantdifferences’’ NoRRs
Seitz&Baron(1990)[20] SAB Rate:exposed=14%;otherwork=13%;home=5% Subjectsusedasownunexposed controls,methodsnotwelldescribed.
Stuckeretal.(1990)[21] SAB NoRRforformaldehyde Largenumberswithunknownexposure
precision,RR,andmeta-analysissize(i.e.thenumberofstudies included) [33]. As a consequence, this model assigns greater weight to smaller studies than the fixed effects model, and thereforemay actually be less conservative [36]. Toavoid this problem,weusedthemethodpresentedbyShoreetal.[37]and usedinseveralsubsequentmeta-analyses[38–42].Inthismethod, the summary RR estimate is calculated by directly weighing individual studies by their precision, while between-study heterogeneityisonlyincorporatedintothesummaryRR’svariance (i.e.the95%CI).FunnelplotsandEgger’sandBegg’stestswereused toevaluatepublicationbias[43,44].Missingconfidenceintervals incohortstudieswerecalculatedusingByarsapproximation[45].
All p-values are one-sided since there was a clear a priori hypothesis that formaldehyde would increase, not decrease, reproductiveanddevelopmentaloutcomes.
3.4. Results
Sevenstudieswithdataonmaternalformaldehydeexposure andSABand12studieswithdataonallcombineddevelopmental outcomeswereusedinthismeta-analysis.Inaddition,onestudyof
SABandpaternalformaldehydeexposurewasidentified[12],and thiswasincludedwiththematernalexposurestudiesinseparate analyses.Table1 showsthedatafromallhumanstudies,while Table2showsthosestudiesexcludedfromthemeta-analysisand reasonsforexclusion.Fig.2aandbshowstheForestplotsforthe analysesofSABandcombinedpregnancyoutcomesformaternal formaldehydeexposures,respectively.Ofthestudiesofmaternal formaldehyde exposure, 5 of the 7 (71%) in the SAB analysis (Fig.2a)and9ofthe12(75%)inthealloutcomesanalysis(Fig.2b), hadrelativerisks1.01.
Theresultsofthemeta-analysesareshowninTable3.Inthe meta-analysisof SABandmaternal formaldehydeexposure,the summaryrelativeriskwas1.76(95%CI,1.20–2.59).Thesummary relativerisksforalloutcomescombinedformaternalformalde- hydeexposurewas1.54(95%CI,1.27–1.88).Inanalyseslimited only to those studies that assessed formaldehyde exposure by methodsotherthandirectself-reports,thesummaryrelativerisks forSABandalloutcomescombinedwere1.29(95%CI,0.52–3.21) and1.40(95%CI,1.11–1.78),respectively.Inanalyseslimitedto studiesusing directself-reported formaldehyde exposureinfor- mation, thecorresponding summary relative risks were higher
Overall (I-squared = 33.3%, p = 0.174) Axelsson et al., 1984
John et al., 1994
Taskinen et al., 1999 Hemminki et al., 1982 Hemminki et al., 1985
Taskinen et al., 1994 Saurel-Cubizolles et al., 1994
.25 1 13
Odds ratio
5.6%
(0.88-12.34) 3.29
9.2%
(0.36-2.82) 1.01
11.8%
(0.28-1.73) 0.7
18.4%
(1.0-4.3) 2.1
37.2%
(1.01-2.82) 1.68
7.3%
(1.1-11.2) 3.5
10.5%
(1.2-8.3) 3.2
100%
(1.29-2.41) 1.76
[19]
[17]
[18]
[22]
[23]
[24]
[15]
Weight (%) OR* (95% CI)
Reference Study
[19]
[27]
[26]
[29]
[17]
[18]
[22]
[30]
[23]
[24]
[15]
[28]
OR 95% CI Weight (%) Reference Study
Overall (I-squared = 1.6%, p = 0.429) Hemminki 1982
Taskinen et al., 1999 Ericson et al., 1984 Grazulevicienne et al., 1998
Taskinen et al., 1994 Saurel-Cubizolles et al., 1994 Hemminki 1985 (sep controls) Dulskiene and Grazulevicine, 2005
Maroziene et al., 2002 Axelsson, 1984
Zhu et al., 2006 John et al., 1994
.25 1 13
Odds ratio
2.2%
0.88-12.34 3.29
17.7%
0.81-2.07 1.24
1.8%
0.22-4.18 0.96
21.9%
0.9-2.09 1.37
3.7%
0.36-2.82 1.01
6.5%
0.4-1.87 0.87
7.3%
1.0-4.3 2.1
7.7%
1.03-4.26 2.09
14.8%
1.01-2.82 1.68
2.9%
1.1-11.2 3.5
4.2%
1.2-8.3 3.2
9.4%
0.8-2.9 1.5
100%
1.27-1.88 1.54
(a)
(b)
Fig.2.Forestplotforstudiesofspontaneousabortion(SAB)andallreproductiveoutcomescombined.TheseForestplotsshowthatORsequaltoorabove1.01werefoundin (a)5ofthe7(71%)studiesintheSABanalysis,and(b)9ofthe12(75%)studiesinthealloutcomesanalysis.Mostoftheconfidenceintervalsinthealloutcomesanalysiswere wellabove1,indicatinghighersignificance.*ORsin(a)werecalculatedfromSABdatareportedinAxelssonetal.(1984),Hemminkietal.(1982),andSaurel-Cubizzolesetal.
(1994),andrecalculatedfromthedataprovidedinHemminkietal.(1985)asdescribedinthefootnotetoTable1.ORsin(b)werecalculatedfromdataoncongenital malformationsreportedinEricsonetal.(1984),andfromdataonSABandcongenitalmalformationscombinedinHemminkietal.(1985)astheseoutcomeswerebasedon separatecontrols.
A.Duongetal./MutationResearch728(2011)118–138 126
(SAB:RR=2.04(95%CI,1.40–2.97);alloutcomes:RR=1.95(95%
CI,1.35–2.81)).
When the Lindbohm et al. study of paternal exposure was addedtothestudiesofmaternalexposurethesummaryrelative risksforSABandalloutcomescombinedwere1.29(95%CI,0.94–
1.76)and1.34(95%CI,1.10–1.62),respectively.
Fig.3showsthefunnelplotsforpublicationbiasforthemeta- analysesof SAB (Fig.3a)and all outcomes combined(Fig.3b).
Withoutpublicationbias,afunnelshapeisexpectedintheseplots
sinceRRsfromlargerstudies,whichhavesmallerstandarderrors (SEs),areexpectedtohavelessdispersionduetorandomchance thantheRRsfromsmallerstudies(whichhavelargerSEs)[44].For maternalformaldehydeexposuresand SAB,obviouspublication biaswasnotseeninthefunnelplot(Fig.3a),orinEggers(p=0.65) orBeggs(p=0.45)tests.Similarly,noclearevidenceofpublication biaswasseeninthemeta-analysisofalloutcomescombinedinthe funnelplot(Fig.3b)orinEggers(p=0.25)orBeggstests(p=0.34).
3.5. Discussion
As a whole, the results of this meta-analysis provide some evidencethatmaternalformaldehydeexposureisassociatedwith SABandpossiblyotherreproductiveoutcomes.SummaryRRswere elevatedinanalysesofSAB(RR=1.76;95%CI,1.20–2.59)andall reproductiveoutcomescombined(RR=1.54;95%CI,1.27–1.88).
Thelowp-valuesassociatedwiththesesummaryRRs(p=0.002
and <0.0001for SABand all outcomescombined,respectively)
showthattheexcessrelativerisksareunlikelyduetochance.For maternalexposure,statisticallysignificantheterogeneitywasnot seen in the meta-analysis of SAB (X2=8.99, p=0.17) or all outcomescombined(X2=11.2,p=0.43).Inaddition,thefactthat greaterthan70%oftheindividualRRsinbothanalyseswereabove 1.0,providessomeindicationofthatthepositiveresultswerefairly consistentacrossstudies.SummaryRRsdecreasedsomewhatfor bothSABandalloutcomescombinedwhenthelargeLindbohm et al. study of paternal formaldehyde exposure was included, though no definite conclusion canbe madebased on only one paternalstudy.
An analysis of dose–response can be an important part of assessingcausalinference,althoughitisnotasinequanon,andin some instances where a true association exists, a clear dose–
response relationship may not bepresent [34]. Six studiesdid provide some dose–response data [12,15,24,28–30]. In the Lindbohm etal. studyof paternalexposures,ORsfor SABwere near 1.0 in both the high (OR=1.0) and low exposure groups (OR=1.1). In several studies,ORs were elevated inthe highest exposuregroup,butnotinthelowerexposuregroups[24,29].For example,Grazulevicieneetal.measuredlowbirthweightriskin3 regionswithdifferentformaldehydeconcentrations,andcrudeRR increased with higher ambient concentration of formaldehyde from1.0(referencegroup)to1.02(95%CI,0.76–1.38)to1.68(95%
CI,1.24–2.27)forexposuregroupsof<1.94
m
g/m3,1.94–3.5m
g/m3, and >3.5
m
g/m3, respectively. In other studies, ORs werehigherintheexposedgroupsthanintheunexposedcontrols,buta clearmonotonicdose–responserelationshipwasnotseen[15,30].
Forexample,inTaskinenetal.,1999,SABORswere1.0,2.4(1.2–
4.8),1.8(0.8–4.0),and3.2(1.2–8.3)intheunexposed(reference), Table3
Resultsofthemeta-analysisofformaldehydeandadversepregnancyoutcomes.
N Fixedeffects Shore Randomeffects Heterogeneity
RR CIL CIU CIL CIU RR CIL CIU X2 p
Maternalexposureonly
Spontaneousabortion 7 1.76 1.29 2.41 1.20 2.59 1.80 1.19 2.70 8.99 0.17
Self-reportedexposuredata 4 2.04 1.40 2.97 –a – – – – 1.89 0.60
Notself-reported 3 1.29 0.74 2.25 0.52 3.21 1.31 0.52 3.25 5.34 0.07
Alloutcomes 12 1.54 1.27 1.88 1.27 1.88 1.55 1.27 1.89 11.18 0.43
Self-reportedexposuredata 5 1.95 1.35 2.81 – – – – – 2.83 0.59
Notself-reported 7 1.40 1.11 1.78 1.11 1.78 1.41 1.11 1.79 6.19 0.40
Maternalandpaternalexposureb
Spontaneousabortion 8 1.29 1.04 1.59 0.94 1.76 1.58 1.06 2.35 16.03 0.02
Alloutcomes 13 1.34 1.14 1.57 1.10 1.62 1.45 1.17 1.80 17.36 0.14
aShoreandrandomeffectsmodelswereonlydonewhenheterogeneitywaspresent.
b IncludesallthematernalexposurestudiesplustheLindbohmetal.,1991studyonpaternalformaldehydeexposure.
0.2.4.6.8SE of log(OR)
-1 0 1 2
log(OR)
Funnel plot with pseudo 95% confidence limits
0.2.4.6.8SE of log(OR)
-1 0 1 2
log(OR)
Funnel plot with pseudo 95% confidence limits (a)
(b)
Fig.3.Funnelplotofstudiesofspontaneousabortion(SAB)andallreproductive outcomescombined.Publicationbiasisnotapparentintheanalysisof(a)SABand (b)alloutcomes.TheSEmeansstandarderror.