Content uploaded by Donald T Wigle
Author content
All content in this area was uploaded by Donald T Wigle on Dec 08, 2015
Content may be subject to copyright.
373
Journal of Toxicology and Environmental Health, Part B, 11:373–517, 2008
Copyright © Taylor & Francis Group, LLC
ISSN: 1093-7404 print / 1521-6950 online
DOI: 10.1080/10937400801921320
EPIDEMIOLOGIC EVIDENCE OF RELATIONSHIPS BETWEEN REPRODUCTIVE AND
CHILD HEALTH OUTCOMES AND ENVIRONMENTAL CHEMICAL CONTAMINANTS
Donald T. Wigle1, Tye E. Arbuckle2, Michelle C. Turner1, Annie Bérubé3, Qiuying Yang4,
Shiliang Liu5, Daniel Krewski1
1McLaughlin Centre for Population Health Risk Assessment, University of Ottawa, Ottawa,
Ontario, 2Healthy Environments and Consumer Safety Branch, Health Canada, Ottawa, Ontario,
3Vulnerable Populations Division, Safe Environments Program, Health Canada, Ottawa, Ontario,
4OMNI Research Group, Department of Obstetrics and Gynecology, Faculty of Medicine,
University of Ottawa and Ottawa Health Research Institute, Ottawa, Ontario, and 5Centre for
Healthy Human Development, Public Health Agency of Canada, Ottawa, Ontario, Canada
This review summarizes the level of epidemiologic evidence for relationships between prenatal and/or early life
exposure to environmental chemical contaminants and fetal, child, and adult health. Discussion focuses on fetal
loss, intrauterine growth restriction, preterm birth, birth defects, respiratory and other childhood diseases, neurop-
sychological deficits, premature or delayed sexual maturation, and certain adult cancers linked to fetal or childhood
exposures. Environmental exposures considered here include chemical toxicants in air, water, soil/house dust and
foods (including human breast milk), and consumer products. Reports reviewed here included original epidemio-
logic studies (with at least basic descriptions of methods and results), literature reviews, expert group reports, meta-
analyses, and pooled analyses. Levels of evidence for causal relationships were categorized as sufficient, limited, or
inadequate according to predefined criteria. There was sufficient epidemiological evidence for causal relationships
between several adverse pregnancy or child health outcomes and prenatal or childhood exposure to environmental
chemical contaminants. These included prenatal high-level methylmercury (CH3Hg) exposure (delayed developmental
milestones and cognitive, motor, auditory, and visual deficits), high-level prenatal exposure to polychlorinated
biphenyls (PCBs), polychlorinated dibenzofurans (PCDFs), and related toxicants (neonatal tooth abnormalities, cog-
nitive and motor deficits), maternal active smoking (delayed conception, preterm birth, fetal growth deficit [FGD]
and sudden infant death syndrome [SIDS]) and prenatal environmental tobacco smoke (ETS) exposure (preterm
birth), low-level childhood lead exposure (cognitive deficits and renal tubular damage), high-level childhood CH3Hg
exposure (visual deficits), high-level childhood exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) (chloracne),
childhood ETS exposure (SIDS, new-onset asthma, increased asthma severity, lung and middle ear infections, and
adult breast and lung cancer), childhood exposure to biomass smoke (lung infections), and childhood exposure to
outdoor air pollutants (increased asthma severity). Evidence for some proven relationships came from investigation
of relatively small numbers of children with high-dose prenatal or early childhood exposures, e.g., CH3Hg poisoning
episodes in Japan and Iraq. In contrast, consensus on a causal relationship between incident asthma and ETS expo-
sure came only recently after many studies and prolonged debate. There were many relationships supported by
limited epidemiologic evidence, ranging from several studies with fairly consistent findings and evidence of dose-
response relationships to those where 20 or more studies provided inconsistent or otherwise less than convincing
evidence of an association. The latter included childhood cancer and parental or childhood exposures to pesticides.
In most cases, relationships supported by inadequate epidemiologic evidence reflect scarcity of evidence as
opposed to strong evidence of no effect. This summary points to three main needs: (1) Where relationships between
child health and environmental exposures are supported by sufficient evidence of causal relationships, there is a
need for (a) policies and programs to minimize population exposures and (b) population-based biomonitoring to
track exposure levels, i.e., through ongoing or periodic surveys with measurements of contaminant levels in blood,
urine and other samples. (2) For relationships supported by limited evidence, there is a need for targeted research
and policy options ranging from ongoing evaluation of evidence to proactive actions. (3) There is a great need for
population-based, multidisciplinary and collaborative research on the many relationships supported by inadequate
evidence, as these represent major knowledge gaps. Expert groups faced with evaluating epidemiologic evidence of
Funding for this review was provided by the National Collaborating Centre for Environmental Health. The authors acknowledge
Robert Cushman for providing helpful comments. Daniel Krewski is the NSERC/SSHRC/McLaughlin Chair in Population Health Risk
Assessment at the University of Ottawa.
This work does not necessarily reflect the views of Health Canada and the Public Health Agency of Canada and no official endorsement
should be inferred. The findings and conclusions of this report are those of the authors and do not necessarily represent the views of
Health Canada and the Public Health Agency of Canada.
Address correspondence to Donald T. Wigle, McLaughlin Centre for Population Health Risk Assessment, University of Ottawa,
Room 318B, One Stewart Street, Ottawa, ON K1N 6N5, Canada. E-mail: don.wigle@sympatico.ca
374 D. T. WIGLE ET AL.
potential causal relationships repeatedly encounter problems in summarizing the available data. A major driver for
undertaking such summaries is the need to compensate for the limited sample sizes of individual epidemiologic
studies. Sample size limitations are major obstacles to exploration of prenatal, paternal, and childhood exposures
during specific time windows, exposure intensity, exposure–exposure or exposure–gene interactions, and relatively
rare health outcomes such as childhood cancer. Such research needs call for investments in research infrastructure,
including human resources and methods development (standardized protocols, biomarker research, validated expo-
sure metrics, reference analytic laboratories). These are needed to generate research findings that can be compared
and subjected to pooled analyses aimed at knowledge synthesis.
Epidemiologic research conducted mainly since 1970 has demonstrated several causal relationships
and many possible associations between parental or childhood environmental exposures and
adverse pregnancy, childhood, and adult health outcomes. Toxicologic research supports assess-
ment of biologic plausibility, one of the major criteria for evaluating cause-effect relationships in
humans since first applied by the U.S. Surgeon General in assessing human health effects of smoking
(U.S. Department of Health, 1964). Epidemiologic studies also draw on toxicologic research
for clues as to possible causes of human diseases and functional deficits and for biomarkers of
exposure, susceptibility, and adverse health effects.
Findings from epidemiologic and toxicologic studies have shown that the risk of adverse preg-
nancy, child, and delayed health outcomes depends not only on the dose and potency of a given
toxicant, but also on the occurrence of exposure during critical developmental time periods (Selevan
et al., 2000). Such evidence of critical exposure windows is congruent with biomolecular research
showing the dependence of fetal and child development on a complex orchestration of genes in
specific cell types at different times (National Academy of Sciences, 2000b). Disruption of prenatal
and early childhood developmental processes can produce permanent disability and functional
deficits, as well as delayed effects such as cancer later in life.
Children may have relatively high exposures to some environmental contaminants because of
their behavior, diet and metabolic and physiologic characteristics (Moya et al., 2004). Hand–mouth
behavior is common among young children and increases exposure to contaminants in soil, house
dust, and toys. Children eat relatively high amounts of certain foods that may contain pesticide
residues or other toxicants and take in more air, water, and food per unit body weight per day than
adults. They also have age-dependent differences in the absorption, distribution, metabolism, and
excretion of chemicals (National Academy of Sciences, 1993). For instance, breast-fed infants have
relatively high exposures to polychlorinated biphenyls (PCBs) and certain other lipid-soluble toxicants
at a time of rapid growth and development (Berlin et al., 2002). Children in disadvantaged households
(low-income and/or low parental education) are at increased risk of exposure to many environmental
hazards, including lead (Bernard & McGeehin 2003), tobacco smoke (Lund & Helgason 2005),
cockroach allergens (Cohn et al., 2006), and outdoor air pollutants (Chaix et al., 2006).
Several recent reports or books provide high-level summaries of current knowledge of relation-
ships between child health and the environment (American Academy of Pediatrics Committee on
Environmental Health, 2003; European Environment Agency and the WHO Regional Office for
Europe [Copenhagen], 2002; Wigle, 2003; Wigle et al., 2006). In this review, we survey the level of
epidemiologic evidence for relationships between adverse reproductive and child health outcomes
and preconceptual, prenatal, and childhood exposures to environmental chemical contaminants.
Wherever possible, we rely on peer-reviewed expert group reports supplemented by subsequently
published original studies. We occasionally cite original studies included in an expert group report
to illustrate the strengths or weaknesses of supporting evidence. Space limitations precluded
comprehensive discussion of each of the many exposure–outcome relationships reviewed here.
Thus our review is a survey of the current state of knowledge in this field, as opposed to a collection of
exhaustive evaluations. Our goal was to categorize the level of evidence for many exposure–outcome
relationships to support public health research and policy planning (further discussed in the Conclusion).
We defined three basic categories for level of evidence (see Methods). Many relationships were
supported by limited evidence; it was beyond the scope of this review to define a fourth category
such as suggestive evidence, as this would have entailed substantial lengthening of the article.
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 375
Discussion is organized by health outcomes, including functional deficits, disability, and structural
abnormalities. These include fetal loss, intrauterine growth restriction, preterm birth, birth defects, can-
cer (including certain adult cancers linked to prenatal or childhood exposures—note that prenatal
exposure is defined throughout the text as prenatal maternal exposure), asthma, other childhood dis-
eases, neuropsychological deficits, and premature or delayed sexual maturation. Environmental expo-
sures considered here include chemical toxicants in air, water, soil, house dust, foods, and consumer
products. Evidence of health effects at high exposure levels is briefly summarized, but the major focus is
on potential health effects at relatively low background exposure levels to which the majority of the
population is likely exposed. The tables summarize key findings for relationships supported by limited
or sufficient epidemiologic evidence (see definitions in Methods). Additional information on the epide-
miologic evidence published up to 2004 for these relationships is available elsewhere (Wigle, 2005).
METHODS
This review included peer-reviewed English-language publications and government reports identi-
fied from PubMed and TOXNET searches and other reports identified from bibliographies of retrieved
articles published by December 31, 2006. Search strategies included key words for various combina-
tions of health outcome and environmental exposure. The PubMed “related articles” function was
used to search for other relevant articles not retrieved in initial key word searches. Key words for
adverse health outcomes included (but were not limited to) fertility, conception, pregnancy, sponta-
neous abortion, stillbirth, fetal death, birth weight, gestation length, preterm birth, birth defect, con-
genital anomaly, chromosomal abnormality, sudden infant death syndrome, asthma, otitis media,
bronchitis, bronchiolitis, pneumonia, allergy, growth, milestone, cognitive, psychomotor, auditory,
visual, hyperactivity, attention, childhood cancer, leukemia, lymphoma, brain cancer, Wilms, neuro-
blastoma, germ cell, bone cancer, sarcoma, retinoblastoma, testicular cancer, breast cancer, lung can-
cer, chloracne, teeth, enamel, dentin, renal, menarche, puberty, and Tanner stage. For exposure, key
words included (but were not limited to) environment, chemical, metal, lead, mercury, cadmium,
manganese, arsenic, polybrominated biphenyl (PBB), polychlorinated biphenyl (PCB), dioxin, pesti-
cide, environmental tobacco smoke, air pollution, smoke, particulate matter, carbon monoxide, poly-
cyclic aromatic hydrocarbon (PAH), ozone, volatile, water, disinfection by-product (DBP), nitrate,
nitrite, bisphenol, and phthalate. Other searches were done using names of authors of relevant arti-
cles. Terms such as case-control, cohort, review, and meta-analysis were used to narrow some
searches. Reviewed reports included original epidemiologic studies (with at least basic descriptions of
methods and results), literature reviews, expert group reports, meta-analyses, and pooled analyses.
When authoritative review articles were available, consideration of original reports was generally lim-
ited to those published since the year before the most recent review was published. Excluded papers
included case reports, analytic studies with fewer than five exposed cases or case parents, studies pub-
lished more than a year before a recent authoritative review, and preliminary reports of subsequently
published studies. All included papers are cited in the text.
Levels of evidence for causal relationships were defined as: (i) sufficient—at least one expert
group has reviewed the available evidence and published a peer-reviewed report indicating a con-
sensus view that there is a causal relationship, (ii) limited—evidence is suggestive of an association
between the agent and the outcome but is limited (and may or may not represent a causal relation-
ship) because chance, bias and confounding cannot be ruled out with confidence, e.g., at least one
high-quality study shows a positive association but the results of other studies are inconsistent and,
(iii) inadequate—available studies are of insufficient quality (e.g., available studies have failed to
adequately control for confounding or have inadequate exposure assessment), consistency or statis-
tical power to permit a conclusion regarding the presence or absence of an association or no studies
exist that examine the relationship. We did not attempt to identify exposure–outcome associations
for which there is limited or sufficient epidemiologic evidence of no causal relationship as the
limitations of available studies precluded firm conclusions about the absence of any risk. The defini-
tions of limited and inadequate epidemiologic evidence are those used recently by the U.S.
National Academy of Sciences (National Academy of Sciences, 2000a)
376 D. T. WIGLE ET AL.
RESULTS
Adverse Pregnancy Outcomes
The level of epidemiologic evidence for associations between delayed conception and fetal
growth and survival and environmental factors is summarized in Table 1.
Delayed Conception (Time to Pregnancy) Couple fecundability is defined as the probability
of conception in a menstrual cycle exposed to unprotected intercourse and can be an indicator of a
wide range of reproductive processes from gametogenesis to survival of the conceptus up to the
time of detection (Baird et al., 1986). The number of menstrual cycles it takes for a couple to con-
ceive (time to pregnancy) can be used to assess fecundability. Delayed conception may be assessed
as a continuous (e.g., number of weeks of unprotected intercourse before conception) or categorical
variable (e.g., duration of unprotected intercourse before conception greater than 6 or 12 mo).
Time to pregnancy studies are most often retrospective; however, prospective studies that recruit
couples at the start of their attempt at pregnancy and incorporate regular testing and follow-up (as
in studies of early pregnancy loss described below) are becoming more feasible (Joffe et al., 2005).
Lead Male partner occupational exposure, inadequate evidence: While sperm concentration
has been found to be reduced among men with a blood lead concentration above 44 μg/dl (Bonde
et al., 2002), results of well-designed time to pregnancy studies have been inconsistent. In a large
multicountry European study, no clear pattern of association of time to pregnancy with short-term
lead exposure, with duration of lead exposure or with cumulative exposure to lead was observed
(Joffe et al., 2003). A more recent study of time to pregnancy among lead battery workers in Taiwan
reported a dose-response relationship with significant decreases in the likelihood of conception at
blood lead levels of 30–39 and ≥40 μg/dl (respective hazard ratios of 0.50, 95% CI 0.34–0.74, and
0.38, 95% CI 0.26–0.56) (Shiau et al., 2004).
Unspecified heavy metals Male partner occupational exposure, inadequate evidence: Studies
of occupational exposures to other heavy metals are limited. A study of workers in a smelter
reported a nonsignificant dose response relation with shorter median waiting times if only one parent
worked in the smelter compared to couples where both were employed in the smelter (Wulff et al.,
1999). Other studies have reported that male welders had lower fecundability ratios than nonweld-
ing metal workers (Hjollund et al., 1998), and suggest that male exposure to metal fumes and
solvents in a mint is associated with an increase in time to pregnancy (Figa-Talamanca et al., 2000).
PCBs Female partner exposure, inadequate evidence:PCB exposure via fish consumption has
been examined as a risk factor for longer time to pregnancy in several cohorts. A recent review of
these studies has concluded that no firm conclusions can be drawn due to uncertainties in the
exposure estimation and inconsistencies in the results (Toft et al., 2004). One study has reported a
weak and inconclusive association between serum levels of PCBs and time to pregnancy (Law et al.,
2005), while another provided no evidence of an adverse effect related to serum PCB–153 levels
(Axmon et al., 2004).
Pesticides Male partner exposure, specific pesticides, inadequate evidence: Studies from
Finland and Canada have examined risks associated with specific pesticides. The only statistically
significant finding was for unprotected use of pyrethroids (fecundability density ratio [FDR]=0.40,
95% CI=0.19–0.85) (Sallmen et al., 2003); however, both studies suggested that fecundability
might be reduced for fungicides (Curtis et al., 1999; Sallmen et al., 2003). One study has suggested
that dichlorodiphenyltrichloroethane (DDT) exposure may be associated with reduced fecundability
(Cocco et al., 2005).
Female partner exposure, specific pesticides, inadequate evidence: Although none of the
results were statistically significant, one study has reported that preconceptual pesticide use was
associated with reduced fecundability for dicamba, glyphosate, 2,4-dichlorophenoxyacetic acid
(2,4-D), organophosphates, and thiocarbamates (Curtis et al., 1999). Two studies have used
biomonitoring data to establish exposure. One study reported that DDT exposure of mothers was
associated with decreased fecundability of their daughters but dichlorodiphenyldichloroethylene
(DDE) exposure was linked with increased fecundability of their daughters (Cohn et al., 2003).
Another study found that only at the highest concentration (≥60 μg/L) was preconceptual serum
377
TABLE 1. Role of Environmental Toxicants in Delayed Conception and Fetal Growth and Survival
Toxicant Exposure Delayed concepaSpont. aborbStillbirthcPreterm birthdFGDe
Lead Prenatal LILL
Paternal I LIII
Mercury Prenatal I I
Arsenic Prenatal LDrinking water—I
Airborne—I
Drinking water—I
Airborne—I
Drinking water—I
Airborne—I
Paternal Airborne—I Airborne—I Airborne—I
Cadmium Prenatal I I I
Other and unspecified metals Paternal I
PCBs Prenatal I High-level—I
Low-level—I
High-level—I
Low-level—I
Environ—I
Occup—I
Environ—I
Occup—I
PBBs Prenatal I
TCDD Prenatal I I
Paternal I I I I
2,4,5-T, chlorophenate wood
preservatives
Paternal I I I I
Other chlorophenoxy herbicides Prenatal I (2,4-D, dicamba) I (2,4-D, 2,4-DB, MCPA) I I
Paternal L (2,4-D, MCPA, dicamba) I I
Other or unspecified herbicides Prenatal I (glyphosate) I (atrazine, glyphosate) I I (atrazine,
metolachlor,
cyanazine)
Paternal I (carbaryl, thiocarbamate,
atrazine, glyphosate)
I (atrazine)
DDT/DDE Prenatal I LLLL
Paternal I I
Organophosphate insecticides Prenatal I I I I I
Paternal I
Other or unspecified
insecticides, repellents
Prenatal I (DEET, organochlorine) I (pyrethroids, organochlorine,
unspecified)
I (organochlorine,
DEET, pyrethroids)
I (DEET, propoxur,
organochlorine)
Paternal I (pyrethroids) I (carbaryl, unspecified insecticdes)
Fungicides (any) Prenatal I I I I
Paternal I I
Ethylene oxide Prenatal L
Paternal I
Unspecified pesticides Prenatal I I I I I
Paternal I I I I I
(Continued)
378
TABLE 1. (Continued)
Toxicant Exposure Delayed concepaSpont. aborbStillbirthcPreterm birthdFGDe
Active smoking Prenatal S LLSS
Paternal I
Environmental tobacco
smoke
Prenatal I LISL
Outdoor air pollutionf
near residence
Prenatal I I LL
Airborne industrial emissions Prenatal II
Drinking water DBPs Prenatal LLIL
Drinking water nitrate Prenatal I I L
Hazardous waste disposal sites Prenatal I I I
Chlorinated solvents Prenatal LII
Glycol ethers Prenatal I L
Paternal I
Other or unspecified solvents Prenatal I LIII
Paternal I LIII
Bisphenol A Prenatal I
Oil, oil products Paternal I
Plastics Paternal I
Note. TCDD = 2,3,7,8-tetrachlorodibenzo-p-dioxin. S, Sufficient evidence—based on peer-reviewed reports of expert groups or authoritative reviews that concluded that a causal rela-
tionship existed. L, Limited evidence—includes relationships for which several epidemiologic studies, including at least one case-control or cohort study, found fairly consistent associations
and evidence of exposure-risk relationships after control for potential confounders. I, Inadequate evidence—relationships for which epidemiologic studies were limited in number and quality
(e.g., small studies, ecologic studies, limited control of potential confounders), had inconsistent results, or found little or no evidence of exposure-risk relationships.
aFor delayed conception, prenatal or paternal exposure, respectively, refer to preconceptual female or male partner exposure.
bClinically apparent pregnancy loss before gestation week 20.
cCurrently defined as fetal death after gestation week 20; previously defined as fetal death after gestation week 28.
dGestation length <37 wk.
eFetal growth deficit comprises small for gestational age (birth weight below 10th percentile based on gestation length) and term low birth weight (birth weight <2500 g for infants born at
37 or more weeks gestation).
fMajor pollutants from fossil fuel combustion.
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 379
DDE concentration associated with reduced fecundability (FDR= 0.65, 95% CI 0.32–1.31) (Law
et al., 2005).
Female partner occupations, unspecified pesticides, inadequate evidence: Two studies have
reported that work in flower production was associated with impaired fecundability, but did not
attempt to determine risks for specific pesticides (Abell et al., 2000; Idrovo et al., 2005).
Male partner occupations, unspecified pesticides, inadequate evidence: A number of European
studies have investigated the role of occupations involving pesticide use on time to pregnancy with
conflicting results (de Cock et al., 1994; Larsen et al., 1998; Petrelli & Figa-Talamanca 2001;
Thonneau et al., 1999). While none of these studies examined risk associated with specific pesti-
cides, some studies attempted to qualify exposures by employing crude measures of intensity of
exposure (e.g., use of protective equipment, duration of exposure) with higher exposures associated
with longer time to pregnancy (de Cock et al., 1994; Petrelli & Figa-Talamanca 2001; Sallmen
et al., 2003).
Tobacco smoke Female partner, active smoking, sufficient evidence: Female smokers have a
dose-related increased risk of delayed conception (Bolumar et al., 1996; Curtis et al., 1997; Hassan &
Killick 2004). The U.S. Surgeon General concluded there was sufficient evidence of a causal rela-
tionship between female partner active smoking and reduced fertility (U.S. Department of Health
and Human Services, 2004).
Female partner ETS exposure, inadequate evidence: There have been few epidemiologic studies
of the role of environmental tobacco smoke (ETS) exposure in delayed conception. The U.S.
Surgeon General reviewed the four available studies and found inadequate evidence for an associ-
ation between reduced female fertility and female partner ETS exposure alone or in combination
with active smoking (U.S. Department of Health and Human Services, 2006). In the Avon Longitu-
dinal Study of Pregnancy and Childhood Study, however, delayed conception was associated with
prenatal active smoking (OR=1.23, 95% CI 0.98–1.49, for delay of over 6 mo; OR =1.54, 95% CI
1.19–2.01, for delay of over 12 mo) and, independently, with prenatal ETS exposure (OR=1.17,
95% CI 1.02–1.37 and OR= 1.14, 95% CI 0.92–1.42, respectively) (Hull et al., 2000).
No consistent pattern has been observed between male partner smoking and fecundability;
it appears that no published studies have assessed male partner ETS exposure and delayed
conception.
Outdoor air pollution Male or female partner exposure, inadequate evidence: One study has
reported an association between ambient sulfur dioxide (SO2) levels and fecundability in the first
unprotected menstrual cycle (Dejmek et al., 2000).
Indoor air pollution Female partner exposure, inadequate evidence: Frequent occupational
exposure to nitrous oxide for midwives has been associated with longer time to pregnancy (FDR = 0.64,
95% CI 0.44–0.95) (Ahlborg et al., 1996).
Solvents Female partner exposure, glycol ethers, inadequate evidence: There are some data
showing that females occupationally exposed to ethylene glycol ethers have longer time to preg-
nancy (FDR = 0.59, 95% CI 0.37–0.94) (Chen et al., 2002b).
Female partner exposure, organic solvents, inadequate evidence: Decreased fecundity was associ-
ated with female partner solvent exposure (FDR = 0.79, 95% CI 0.68–0.93), particularly those using
acetone (FDR=0.72, 95% CI 0.53–0.97) (Wennborg et al., 2001). One study reported that daily tolu-
ene exposure was associated with reduced fecundity in women (FDR = 0.47, 95% CI 0.29–0.77)
(Plenge-Bonig & Karmaus 1999). Similarly, female exposure to formaldehyde at work may have an
adverse effect on fecundity (FDR = 0.64, 95% CI 0.43–0.92) (Taskinen et al., 1999). The Agricultural
Health Study (AHS) cohort, a large study of licensed pesticide applicators in Iowa and North Carolina,
reported that the likelihood of not becoming pregnant after 12 mo of unprotected intercourse was ele-
vated for women (OR=1.42, 95% CI 1.15–1.75) with at least monthly exposure to solvents (Sallmen
et al., 2006). Stronger associations were apparent when solvent exposure was defined as either partner
(OR=1.62, 95% CI 1.20–2.17) or both partners (OR=2.10, 95% CI 1.22–3.60).
Male partner exposure, organic solvents, inadequate evidence: Men with frequent occupational
solvent exposure experienced decreased fecundity (FDR = 0.80, 95% CI 0.57–1.11) (Sallmen et al.,
1998), but studies reported no effect (Luderer et al., 2004; Spinelli et al., 1997). Toluene exposure
380 D. T. WIGLE ET AL.
was not associated with reduced fecundity in men (Plenge-Bonig & Karmaus 1999). The Agricultural
Health Study (AHS) cohort, a large study of licensed pesticide applicators in Iowa and North Carolina,
reported that the likelihood of not becoming pregnant after 12 mo of unprotected intercourse
was elevated for men (OR=1.21, 95% CI 0.93–1.57) with at least monthly exposure to solvents
(Sallmen et al., 2006). Stronger associations were apparent when solvent exposure was defined as
either partner (OR=1.62, 95% CI 1.20–2.17) or both partners (OR=2.10, 95% CI 1.22–3.60).
Other toxicants Male partner exposure, plastics, inadequate evidence: One study reported no
association between workers highly exposed to di(2-ethylhexyl)phthalate and time to pregnancy
(Modigh et al., 2002). While reduced fecundity (FDR = 0.79, 95% CI 0.59–1.05) was observed in
styrene-exposed workers, no dose response was observed in relation to tasks indicating higher
exposure (Kolstad et al., 2000).
Male partner exposure, oil and oil products, inadequate evidence: One study has reported no
major influence of exposure to hydrocarbons on time to conception (Bull et al., 1999).
Summary Except for female partner active smoking (sufficient evidence), epidemiologic
evidence for the role of environmental toxicants in delayed conception is inadequate.
Early Pregnancy Loss Early pregnancy loss is defined as pregnancy detected by daily urinary
human chorionic gonadotrophin (hCG) monitoring with loss occurring less than 6 wk after onset of
the last normal menstrual period (LNMP). Early pregnancy loss occurs in about 10–20% of concep-
tions and is usually not recognized or reported (Ellish et al., 1996; Hjollund et al., 2000; Wang et al.,
2003b). As there have been very few studies of early pregnancy loss and environmental exposures,
this field remains largely unexplored.
Pesticides Maternal exposure, inadequate evidence: A cohort study of newly married Chinese
female textile workers monitored for conception using daily urinary hCG measurements revealed a
dose-response relationship between early pregnancy loss (gestation length <6 wk, confirmed by
hCG) and maternal preconceptual serum DDT levels (2nd vs. 1st tertile, OR=1.07, 95% CI 0.58–1.99;
3rd vs. 1st tertile, OR=1.71, 95% CI 0.93–3.12, p-trend=.06); there were similar results when
serum DDE was modeled (Venners et al., 2005).
Spontaneous Abortion (Gestation Week <20) Spontaneous abortion is defined as uninten-
tional fetal loss before gestation week (GW) 20. This is the definition used in most recent epidemiologic
studies, but some studies have included fetal deaths up to GW 24 or 28. Because early pregnancy loss is
rarely recognized and reported, clinically recognized spontaneous abortions virtually all occur after GW 6.
Lead Maternal exposure, limited evidence: A recent review concluded that high-level prenatal
occupational lead exposure during the 19th and early 20th centuries likely increased the risk of
spontaneous abortion and that limited evidence supports an association at prenatal blood lead
levels below approximately 30 μg/dl (Hertz-Picciotto 2000). In particular, a birth cohort study in
Mexico City revealed a monotonic dose-response relationship between spontaneous abortion and
prenatal blood lead (respective odds ratios for 5–9, 10–14, and ≥15 μg/dl were 2.3, 5.4, and 12.2,
p-trend=.03) (Borja-Aburto et al., 1999).
Paternal occupational exposure, limited evidence: Reviewers noted limited evidence for an
association between spontaneous abortion and paternal occupational lead exposure (Anttila &
Sallmen, 1995; Bellinger, 2005). Spontaneous abortion was associated with preconceptual blood
lead levels of at least 39 μg/dl in a Finnish cohort of occupationally exposed men (OR= 3.0, 95% CI
1.0–8.7) (Lindbohm et al., 1991b). A similar study in British Columbia found no association with
paternal blood lead levels; selection bias is possible as only 38% of workers participated in this
study (Alexander et al., 1996). A retrospective cohort study of Norwegian men reported an elevated
risk of 2nd trimester spontaneous abortion and likely lead exposure based on job titles (OR=2.4,
95% CI 0.8–6.9) (Kristensen et al., 1993).
Inorganic and elemental mercury Maternal exposure, inadequate evidence: Case-control
studies in Massachusetts revealed an association of borderline statistical significance between spon-
taneous abortion and detectable drinking water mercury levels in the community of prenatal resi-
dence (OR=1.5, 95% CI 1.0–2.3) (Aschengrau et al., 1989). In a small retrospective cohort study,
spontaneous abortion risk was not elevated among women occupationally exposed to elemental
mercury (OR=1.07, 95% CI 0.27–4.56, calculated from data in paper) (Elghany et al., 1997).
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 381
Inorganic arsenic Maternal exposure, limited evidence: A case-control study In Massachusetts
observed no association between spontaneous abortion and maternal residence in communities
with detectable arsenic in drinking water supplies (≥1.4 μg/L vs. undetectable, OR=1.5, 95% CI
0.4–4.7); the highest level detected was 1.9 μg/L, well below the current U.S. Environmental
Protection Agency (EPA) drinking water arsenic maximum contaminant level (10 μg/L) (Aschengrau
et al., 1989). A retrospective cohort study in Bangladesh revealed an increased risk of spontaneous
abortion in a town with average drinking-water arsenic levels of 240 μg/L, relative to a comparison
town with arsenic levels below 20 μg/L (RR= 2.82, 95% CI 1.12–7.36, calculated from data in
report) (Ahmad et al., 2001). Spontaneous abortion risk was slightly elevated among women working
in a Swedish copper smelter (OR=1.33, 95% CI 0.94–2.08) (Wulff et al., 2002). Reviewers found
limited epidemiologic evidence and sufficient toxicologic evidence of fetal deaths after prenatal
arsenic exposure (Golub et al., 1998).
PCBs PCB congeners have half-lives in humans or monkeys of about 3–20 yr (Masuda, 2001;
Mes et al., 1995). As their concentrations per unit weight of lipid in maternal or cord blood, breast
milk, or adipose tissue samples are highly correlated, PCB levels in any of these samples provide an
index of prenatal and fetal exposure. For breast-fed infants, they also reflect lactational exposure,
especially when combined with duration of breast feeding. Although most epidemiologic studies
report serum or plasma PCB concentrations adjusted for lipid content (based on total cholesterol
and triglycerides), a substantial proportion of serum organochlorines are not associated with lipid
(Longnecker et al., 2002). Thus use of lipid-adjusted serum PCB levels may contribute to misclassifi-
cation of exposure levels and reduce the chance of observing true associations with health outcomes.
Maternal high-level exposure, inadequate evidence: A retrospective cohort study of women
who consumed cooking oil contaminated by high levels of PCBs, polychlorinated dibenzofurans
(PCDFs) and related toxicants during the Yucheng incident in Taiwan revealed no association
between spontaneous abortion and maternal preconceptual serum PCB levels (baseline maternal
serum PCB >46 vs. ≤46 μg/g lipid, crude OR=1.12, 95% CI 0.34–3.70) (Yu et al., 2000).
Maternal low-level exposure, inadequate evidence: Reviewers found inadequate evidence
for an association between spontaneous abortion and background environmental PCB exposure
(Longnecker et al., 1997). In a Swedish retrospective cohort study of fishing families, spontaneous
abortion was not associated with residence in a region with fish contaminated by relatively high
PCB concentrations (1st trimester fetal death, OR=0.51, 95% CI 0.27–0.96; 2nd trimester,
OR=0.90, 95% CI 0.44–1.83) (Axmon et al., 2000). In the absence of body-burden data, interpre-
tation of these results is difficult. A case-control study nested within a cohort of Chinese textile workers
revealed no association between spontaneous abortion and prenatal serum PCB (per 1 ng/100 g serum
increment, OR= 0.96, 95% CI 0.87–1.05) (Korrick et al., 2001). In a small Japanese case-control study,
spontaneous abortion was not associated with prenatal serum PCB concentration (mean serum PCB,
cases vs. controls, 263.7 ± 136.9 vs. 319.9 ± 189.7 ng/g lipid) (Sugiura-Ogasawara et al., 2003). In an
Australian birth cohort study, pregnancy loss (spontaneous abortion or stillbirth) was not associated with
breast milk PCB levels (≤50 μg/kg lipid vs. undetectable, OR=0.60, 95% CI 0.17–2.14; >50 μg/kg,
OR=1.07, 95% CI 0.34–3.35, p-trend=0.65) (Khanjani & Sim, 2007).
PBBs Maternal exposure, inadequate evidence: A retrospective cohort study of women in
Michigan who ate meat from livestock accidentally poisoned by polybrominated biphenyls (PBBs)
revealed no association between spontaneous abortion and maternal serum PBBs at baseline soon
after exposure (>2 ppb vs. <LD, OR=0.73, 95% CI 0.47–1.13) (Small et al., 2007).
TCDD Maternal environmental exposure, inadequate evidence: A cohort study of women
living in Seveso at the time of the 1976 factory explosion that released substantial amounts of
2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) found no association between spontaneous abortion
and maternal preconceptual serum TCDD levels (per 10-fold increment of maternal serum TCDD,
OR=1.0, 95% CI 0.6–1.6) (Eskenazi et al., 2003).
Paternal occupational exposure, inadequate evidence: Reviewers found insufficient evidence
for an association between spontaneous abortion and paternal occupational exposure to phenoxy
herbicides potentially contaminated with TCDD (National Academy of Sciences, 2003). This review
focused mainly on health risks for Vietnam veterans potentially exposed to Agent Orange (a 50:50
382 D. T. WIGLE ET AL.
mixture of 2,4-D and 2,4,5-trichlorophenoxyacetic acid (2,4,5-T) known to be contaminated with
TCDD). A retrospective cohort study of veterans revealed no association between spontaneous
abortion and paternal serum TCDD levels defined as background (current serum TCDD < 10 pg/g
lipid), low (current serum TCDD ≥ 10 and baseline level 10–109 pg/g lipid) and high (current serum
TCDD ≥ 10 and baseline level ≥ 110 pg/g lipid) (Wolfe et al., 1995). The respective odds ratios for
spontaneous abortion, calculated from data in the paper, were background (OR=1.13, 95% CI
0.81–1.59), low (OR= 1.32, 95% CI 0.94–1.86) and high (OR=0.99, 95% CI 0.68–1.43). A subse-
quent study of wives of men highly exposed to TCDD during production of Agent Orange revealed
no association between spontaneous abortion and paternal serum TCDD (per 10-fold increment of
paternal serum TCDD at conception, OR=0.97, 95% CI 0.88–1.09) (Schnorr et al., 2001).
Pesticides Reviewers noted limited but somewhat inconsistent evidence for associations
between fetal deaths (spontaneous abortions or stillbirths) and maternal or paternal pesticide
exposure indices (Arbuckle & Sever, 1998; Sever et al., 1997). They noted methodologic issues,
especially inadequate exposure assessment and limited statistical power of epidemiologic studies.
A more recent review examined evidence from epidemiologic studies of pregnancy outcome that
assessed pesticide class, family, and/or active ingredient (Weselak et al., 2006). These reviewers
found limited epidemiologic evidence for associations between fetal death and DDT and inade-
quate evidence for associations with other pesticide categories including chlorophenoxy herbicides,
triazine herbicides, or thiocarbamate fungicides.
Maternal exposure, chlorophenoxy herbicides, inadequate evidence: Chlorophenoxy herbicides
comprise many closely related chemical entities including 2,4-dichlorophenoxyacetic acid (2,4-D1),
2,4-dichlorophenoxybutyric acid (2,4-DB), 2-methyl–4-chlorophenoxyacetic acid (MCPA), and 2-
methyl–4-chlorophenoxypropionic acid (MCPP, mecoprop) (Wood, 2007). The Ontario farm
family study reported no association between spontaneous abortion at GW <20 and prenatal farm
use of any chlorophenoxy herbicide (OR=1.1, 95% CI 0.6–2.1) or subtypes including
2,4-D (OR=1.0, 95% CI 0.5–2.0), 2,4-DB (OR=0.7, 95% CI 0.3–1.7) or MCPA (OR=0.9, 95%
CI 0.4–2.0) (Arbuckle et al., 1999). The Ontario study also found no association between these
herbicides and spontaneous abortions before GW 12 or during GW 12–19. A subsequent report of
this study confirmed no association between prenatal farm use of chlorophenoxy herbicides and
spontaneous abortion at GW <12 (OR=0.6, 95% CI 0.4–1.0) or GW 12–19 (OR=1.3, 95% CI
0.8–2.0) (Arbuckle et al., 2001). However, there were associations of borderline statistical significance
between late spontaneous abortion (GW 12–19) and prenatal farm use of 2,4-D (crude OR=1.6,
95% CI 0.9–2.7) or dicamba (crude OR=1.6, 95% CI 0.8–3.2). A case-only analysis within this study
showed that early spontaneous abortion (GW <12) was more likely after preconceptual compared to
postconceptual (prenatal) chlorophenoxy herbicide use (OR=1.9, 95% CI 1.1–3.3). In the Ontario
study, pesticide exposure reflected both pesticide use by the farm operator spouse (80% males) and
indirect exposure of the other spouse since the study was limited to couples on operating farms.
Maternal exposure, other herbicides, inadequate evidence: The Ontario farm family study
reported no association between spontaneous abortion before GW 20 and prenatal farm use of
atrazine (OR= 0.8, 95% CI 0.5–1.2) or glyphosate (OR =1.1, 95% CI 0.7–1.7) (Arbuckle et al.,
2001). There was also no association between these pesticides and spontaneous abortions before
GW 12 or during GW 12–19.
Maternal exposure, DDT/DDE, limited evidence: A recent review found limited evidence
based on five studies that reported associations between fetal death and biomarkers of prenatal
DDT/DDE exposure (Weselak et al., 2006). Among recent studies cited in their review, a large U.S.
nation-wide study found a nonmonotonic dose-response relationship between fetal death at any
gestation length and prenatal serum DDE (maternal serum DDE increment of 60 μg/L, OR=1.4,
95% CI 1.1–1.6) but not DDT levels (Longnecker et al., 2005). Among newly married Chinese
female textile workers monitored for conception using daily urinary hCG measurements, there was
no association between clinical spontaneous abortion (GW 6–19) and maternal preconceptual
serum DDT levels (2nd vs. 1st tertile, OR=1.22, 95% CI 0.51–2.92; 3rd vs. 1st tertile, OR= 1.28,
1First used in 1948.
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 383
95% CI 0.53–3.10, p-trend=.61); there were similar results when serum DDE was modeled
(Venners et al., 2005). In an Australian retrospective cohort study, fetal death (any gestation length)
was not associated with breast milk DDT or DDE levels (2nd vs. 1st tertile DDE, OR=0.81, 95% CI
0.0.47–1.42; 3rd vs. 1st tertile, OR=0.76, 95% CI 0.41–1.39) (Khanjani & Sim, 2006).
Maternal exposure, organophosphate insecticides, inadequate evidence: A nested case-control
study in the San Francisco Bay Area revealed a weak association between spontaneous abortion
before GW 28 and prenatal residence in areas treated with aerially applied malathion less than a
week before the outcome (OR=1.20, 95% CI 0.94–1.52) but not for such use 1–4 wk before the out-
come (OR=0.91, 95% CI 0.75–1.12) (Thomas et al., 1992). A retrospective cohort study of Ontario
farm families reported no association between spontaneous abortion (GW <20) and postconceptual
farm use of organophosphate insecticides (OR=0.6, 95% CI 0.4–1.0) (Arbuckle et al., 2001).
Maternal exposure, other specified insecticides or repellents, inadequate evidence: In a
randomized controlled trial of N,N-diethyl-m-toluamide (DEET) to prevent malaria during pregnancy,
spontaneous abortion risk was not increased among exposed women (OR=1.52, 95% CI 0.49–4.85)
(McGready et al., 2001). In an Australian retrospective cohort study, fetal death (any gestation
length) was not associated with breast milk dieldrin, heptachlor epoxide or oxychlordane levels
(Khanjani & Sim. 2006).
Maternal exposure, fungicides, inadequate evidence: Hexachlorobenzene (HCB) is an orga-
nochlorine fungicide used as a seed treatment, especially on wheat. A small case-control pilot study
nested within a cohort of Chinese textile workers revealed an association of borderline statistical
significance between spontaneous abortion and prenatal serum HCB (per 1 ng/100 g serum incre-
ment, OR=1.06, 95% CI 1.00–1.14) (Korrick et al., 2001). A small study of Turkish women
poisoned as children (average age 6 yr) by consumption of hexachlorobenzene-contaminated
wheat seed grain (or by breastfeeding if mothers were exposed) revealed an association between spon-
taneous abortion and maternal serum hexachlorobenzene levels (per log HCB (μg/L), β=2.88 ± 0.91,
p< .001) but did not assess or adjust for potential confounders (Jarrell et al., 1998). There was no asso-
ciation between repeated spontaneous abortion (3 or more) and maternal serum HCB levels (mean
HCB ± SD, cases vs. controls, 17.6 ± 10.2 vs. 21.2 ± 10.0 ng/g lipid, p > .05) in a small Japanese case-
control study (Sugiura-Ogasawara et al., 2003). In an Australian retrospective cohort study, fetal death
(any gestation length) was not associated with breast milk HCB levels (2nd vs. 1st tertile, OR=1.03,
95% CI 0.61–1.75; 3rd vs. 1st tertile, OR= 0.49, 95% CI 0.26–0.90) (Khanjani & Sim, 2006). The
Ontario farm family study reported no association between spontaneous abortion (GW <20) and post-
conceptual farm use of the broad class of fungicides (OR=0.8, 95% CI 0.5–1.1) (Arbuckle et al., 2001).
Ethylene oxide Maternal exposure, limited evidence: Ethylene oxide is used as a grain fumigant
but epidemiological studies have assessed occupational exposures in other settings. Among Finnish
nurses, spontaneous abortion was not associated with exposure to anaesthetic gases (Hemminki
et al., 1985). In a retrospective cohort study of most recent pregnancies among female dental assis-
tants in California who conceived while working full-time, spontaneous abortion was associated
with self-reported occupational ethylene oxide exposure (OR=2.5, 95% CI 1.0–6.3) and was inde-
pendent of age and exposure to nitrous oxide or preparation of mercury amalgams (Rowland et al.,
1996). A South African retrospective cohort study revealed a strong association between spontane-
ous abortion and occupational exposure to ethylene oxide in hospital sterilizing units while working
full-time during the relevant pregnancy (high vs. low exposure, OR= 20.8, 95% CI 2.1–199); this
estimate was based on only 4 spontaneous abortions among 19 highly exposed pregnancies and 1
among 78 pregnancies of women with relatively low exposure (Gresie-Brusin et al., 2006).
Maternal exposure, unspecified pesticides, inadequate evidence: Reviewers found limited
evidence for associations between spontaneous abortion and maternal employment in agriculture
or other occupational exposure to unspecified pesticides, including dose-response relationships
with number of pesticides used and nonuse of protective equipment (Arbuckle & Sever, 1998).
Recently, a Chinese retrospective cohort study reported that spontaneous abortion was associated
with maternal use of drinking water from sources subject to runoff from pesticide-treated cotton
fields (OR= 1.63, 95% CI 1.11–2.39) but not with occupational pesticide exposure (OR= 0.60, 95%
CI 0.43–0.84) (Cho et al., 1999). In a retrospective cohort study of licensed pesticide applicators in
384 D. T. WIGLE ET AL.
Minnesota, spontaneous abortion was associated with prenatal mixing, loading or application of
pesticides (OR=1.81, 95% CI 1.04–3.12) (Garry et al., 2002). Although suggestive, the heterogene-
ity of exposure indices precludes firm conclusions and more research is needed to examine specific
pesticides, critical exposure windows and dose-response relationships.
Paternal occupational exposure, TCDD-free, inadequate evidence: Reviewers found insuffi-
cient evidence for an association between spontaneous abortion and paternal occupational expo-
sure to chlorophenoxy herbicides potentially contaminated with TCDD (National Academy of
Sciences, 2003). See also discussion of paternal occupational TCDD exposure above.
Paternal occupational exposure, TCDD-contaminated chlorophenoxy herbicides, chlorophenoxy
herbicides, limited evidence: An initial report of the Ontario farm family cohort showed no associa-
tion between spontaneous abortions before GW 20 and preconceptual use (mainly paternal) of 2,4-
D (OR=1.3, 95% CI 0.8–2.1), MCPA (OR=1.1, 95% CI 0.6–1.8) or dicamba (OR=1.1, 95% CI
0.6–2.1) (Savitz et al., 1997a). Further analysis revealed an association between preconceptual agri-
cultural use (mainly paternal) of chlorophenoxy herbicides and spontaneous abortion before GW
12 (OR=2.5, 95% CI 1.0–6.4) but not with those at GW 12–19 (OR =0.4, 95% CI 0.2–1.0)
(Arbuckle et al., 1999). The Ontario study also revealed statistically nonsignificant elevated risks of
early spontaneous abortion (GW < 12) related to preconceptual paternal use of 2,4-D (OR=1.9,
95% CI 0.7–4.8) or MCPA (OR=2.3, 95% CI 0.8–6.5). A subsequent report indicated that sponta-
neous abortion at GW < 12 was associated with preconceptual chlorophenoxy herbicide use
(crude OR=1.5, 95% CI 1.1–2.1) and weakly with 2,4-D use (crude OR=1.3, 95% CI 0.9–2.0)
(Arbuckle et al., 2001). Although there was no association between spontaneous abortion at GW
<20 and preconceptual farm use of dicamba (OR=1.1, 95% CI 0.7–1.9, 95% CI 0.7–1.7), there
was a statistically nonsignificant elevated risk of early spontaneous abortion (GW < 12) related to
such exposure (OR= 1.6, 95% CI 0.8–3.2). In a case-only analysis, early spontaneous abortions
(GW < 12) were more likely after preconceptual compared to postconceptual exposure chlorophe-
noxy herbicide use (OR=3.1, 95% CI 1.4–6.4) (Arbuckle et al., 2001). A retrospective cohort study
in Minnesota revealed a statistically nonsignificant association between spontaneous abortion
before GW 13 among the subset of pregnancies conceived during the spring spray season and use
of any chlorophenoxy herbicide (relative to use of any other pesticides, OR= 1.59, 95% CI 0.77–
3.27) (Garry et al., 2002). However, there was an association between spontaneous abortions con-
ceived during spring and paternal combined use of chlorophenoxy, sulfonylurea, and benzothioda-
zole herbicides (relative to use of any other pesticides, OR=2.94, 95% CI 1.40–6.16). In
Minnesota, herbicides are generally applied to crops in spring, and insecticides in summer and fun-
gicides as needed in summer and fall (Garry et al., 2002).
Paternal occupational exposure, other herbicides, inadequate evidence: The Ontario farm
family study reported that spontaneous abortion before GW 20 was associated with preconceptual
use (mainly paternal) of carbaryl (OR=1.9, 95% CI 1.1–3.1) and thiocarbamate crop herbicides
(OR=1.9, 95% CI 1.1–3.3); there were also associations of borderline statistical significance with
atrazine (OR=1.5, 95% CI 0.9–2.4) and glyphosates (OR = 1.5, 95% CI 0.8–2.7) (Savitz et al.,
1997a). A subsequent report of the Ontario study revealed no association between spontaneous
abortion at GW <20 and preconceptual farm use of atrazine (OR=0.8, 95% CI 0.5–1.2) or glypho-
sate (OR=1.1, 95% CI 0.7–1.7) (Arbuckle et al., 2001). The latter report indicated a statistically
nonsignificant elevated risk of early spontaneous abortion (GW <12) related to preconceptual farm
use of glyphosate (OR=1.4, 95% CI 0.8–2.5). The Minnesota retrospective cohort study revealed
associations between spontaneous abortion before GW 13 among the subset of pregnancies
conceived during the spring spray season and use of herbicides including sulfonylurea (relative to
use of any other pesticides, OR=2.11, 95% CI 1.09–4.09) and imidizolinone (OR=2.56, 95%
CI 1.11–5.87) (Garry et al., 2002). Further studies are needed to assess specific herbicides, critical
exposure windows, and dose-response relationships.
Paternal occupational exposure, DDT/DDE, inadequate evidence: Spontaneous abortion
before GW 20 was associated with paternal employment using backpacks to apply DDT and other
insecticides on cotton crops in India (exposed vs. unexposed men, RR= 2.00, 95% CI 1.77–2.26)
(Rupa et al., 1991). A retrospective cohort study of male malaria control workers in Mexico
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 385
reported a weak and statistically nonsignificant association between spontaneous abortion and
paternal DDE body burden estimated from self-reported information on timing, intensity, and dura-
tion of DDT exposure (4th vs. 1st quartile estimated paternal DDE, OR=1.24, 95% CI 0.91–1.70)
(Salazar-Garcia et al., 2004).
Paternal occupational exposure, organophosphate insecticides, inadequate evidence: The
Ontario farm family study reported that spontaneous abortion before GW 20 was not associated
with preconceptual use (mainly paternal) of organophosphate crop insecticides (OR=1.3, 95%
CI 0.7–2.3) (Savitz et al., 1997a). A more recent report of this study also revealed no association
between preconceptual organophosphate insecticide use and early (GW <12) (OR= 1.0, 95%
CI 0.6–1.6) or late (GW 12–19) spontaneous abortion (OR=1.0, 95% CI 0.6–1.7) (Arbuckle et al.,
2001). In a case-only analysis, the Ontario study showed that early spontaneous abortion was asso-
ciated with preconceptual organophosphate insecticide use (compared to postconceptual use,
OR=3.8, 95% CI 1.1–13.4). The significance of the latter finding is not clear but supports the need
for research to clarify the role of preconceptual organophosphate insecticide exposure.
Paternal occupational exposure, other or unspecified insecticides, inadequate evidence: The
Ontario farm family study reported that spontaneous abortion before GW 20 was associated with pre-
conceptual carbaryl use (OR= 2.1, 95% CI 1.1–4.1) (Savitz et al., 1997a). In a large retrospective cohort
of licensed pesticide applicators in Minnesota, spontaneous abortion was not associated with insecticide
use (insecticide and herbicide vs. herbicide only use, OR=1.27, 95% CI 0.68–2.36) (Garry et al., 2002).
Paternal occupational exposure, fungicides, inadequate evidence: The Ontario farm family
study reported that spontaneous abortion before GW 20 was not associated with preconceptual use
(mainly paternal) of fungicides (OR=1.2, 95% CI 0.7–2.1) (Savitz et al., 1997a). In a subsequent
report, there were statistically nonsignificant associations between preconceptual fungicide use and
spontaneous abortion before GW 12 (OR=1.3, 95% CI 0.9–1.9) or during GW 12–19 (OR=1.4,
95% CI 0.9–2.1) (Arbuckle et al., 2001). In a case-only analysis, spontaneous abortion before GW
12 (but not those at GW 12–19) was associated with preconceptual farm fungicide use (compared
to postconceptual use, OR=3.9, 95% CI 1.4–10.3). The Minnesota study revealed associations
between spontaneous abortion and application of the fungicides mancozeb and/or maneb
(compared to unexposed men, OR=1.77, 95% CI 1.11–2.83) and organotin fungicides
(OR=1.55, 95% CI 1.01–2.37) (Garry et al., 2002).
Ethylene oxide Paternal exposure, inadequate evidence: In a Finnish retrospective cohort
study based on linkage of national databases on pregnancy outcome and census information, spon-
taneous abortion was associated with paternal occupations likely exposed to ethylene oxide (based
on job-exposure matrix) (OR=4.7, 95% CI 1.2–18.4) (Lindbohm et al., 1991a).
Paternal occupational exposure, unspecified pesticides, inadequate evidence: Retrospective
cohort studies reported associations between spontaneous abortion and paternal occupational
exposure to unspecified pesticides in semi-enclosed Columbian greenhouses (exposure vs. preex-
posure period, OR=1.79, 95% CI 1.16–2.77) (Restrepo et al., 1990b) and backpack application of
multiple pesticides in India (OR=2.00, 95% CI 1.82–2.66) (Rupa et al., 1991). In another
retrospective cohort study, Norwegian farm families had an elevated risk of spontaneous abortion
(GW 16–27) compared to nonfarm families living in agricultural communities (OR =2.18, 95% CI
1.67–2.85) (Kristensen et al., 1997a). This study could not distinguish between maternal and pater-
nal pesticide exposure; 57% of men and 34% of women worked at least 500 h/yr on their farms. A
small retrospective cohort study in Italy reported an association between spontaneous abortion and
paternal occupation as a pesticide applicator (OR=3.8, 95% CI 1.2–12.0) (Petrelli et al., 2000). In
a cohort of Italian greenhouse workers, there was a strong association between spontaneous abor-
tion and paternal occupational pesticide exposure (compared to unexposed greenhouse workers,
crude OR=3.70, 95% CI 1.04–13.3, calculated from data in paper) (Petrelli et al., 2003). A retro-
spective cohort study of wives of agricultural pesticide aerial applicator pilots revealed no increased
risk of spontaneous abortion (crude OR=0.65, 95% CI 0.41–1.03, calculated from data in paper)
(Roan et al., 1984). Although the collective evidence from these studies is suggestive, the heteroge-
neity of exposure indices precludes firm conclusions.
386 D. T. WIGLE ET AL.
Tobacco smoke Prenatal active smoking, limited evidence: The U.S. Surgeon General
concluded that there is suggestive evidence of a causal relationship between spontaneous abortion
and maternal active smoking (U.S. Department of Health and Human Services, 2001).
Prenatal ETS exposure, limited evidence: Reviewers concluded that the three available epide-
miologic studies provided limited evidence of an association between spontaneous abortion and
prenatal ETS exposure (Lindbohm et al., 2002). An expert panel found limited evidence of a causal
association between spontaneous abortion and prenatal ETS exposure independent of paternal
smoking (California Environmental Protection Agency, 2005). The panel noted limited evidence of
a causal association between spontaneous abortion and paternal active smoking (an important source
of prenatal ETS exposure) but noted that this could arise from preconceptual paternal germ-cell muta-
tions rather than postconceptual maternal ETS exposure. The U.S. Surgeon General reviewed five
available studies and concluded that there was inconsistent evidence of an association between
spontaneous abortion and prenatal ETS exposure (U.S. Department of Health and Human Services,
2006). A subsequently published original case-control study of spontaneous abortions at GW 6–12
revealed an association with maternal ETS exposure (OR=1.67, 95% CI 1.17–2.38) (George et al.,
2006). This study was unusually robust in that both the outcome and exposure were confirmed
using biomarkers (hCG to confirm pregnancy and cotinine to verify exposure status).
Drinking-water disinfection by-products Most epidemiologic studies of DBPs and potential
adverse health outcomes have used drinking-water trihalomethane (THM) concentrations as an
indicator of total disinfection by-product (DBP) concentrations in chlorinated drinking water. DBP
concentrations tend to be highest in chlorinated surface waters that contain relatively high amounts
of natural organic material that reacts with chlorine to generate THMs and many other by-products.
The mutagenic activity of raw drinking water is generally very low but may increase substantially
after chlorination.
Maternal exposure, limited evidence: Reviewers noted limited and fairly consistent evidence for
an association between spontaneous abortion and drinking water disinfection by-products (Bove et al.,
2002; Graves et al., 2001; Nieuwenhuijsen et al., 2000). In experimental animals, high-dose prenatal
exposure to chloroform, bromodichloromethane (BDCM), haloacetonitriles, or haloacetic acids
(HAAs) caused fetal toxicity, including fetal resorptions and reduced fetal weight and survival
(Graves et al., 2001; Nieuwenhuijsen et al., 2000).
Hazardous waste disposal sites Maternal exposure, inadequate evidence: Reviewers found
inadequate evidence for an association between fetal deaths and prenatal residential proximity to
hazardous waste landfill sites or incinerators; there were few extant studies of spontaneous abortion
and such exposure (Johnson & DeRosa, 1995; Vrijheid, 2000). Subsequently published studies
revealed no association between spontaneous abortion and maternal residential proximity to
high-dioxin emission incinerators (OR=0.82, 95% CI 0.54–1.20, p-trend=.97 (over ten 1-km
increments)) (Tango et al., 2004) or hazardous waste disposal sites (high vs. low/moderate hazard
sites, OR=0.75, 95% CI 0.28–1.99) (Gilbreath & Kass, 2006a).
Solvents Prenatal occupational exposure, chlorinated solvents, limited evidence: Spontaneous
abortions were associated with prenatal occupational exposure to tetrachloroethylene in Scandinavia
(≥1 h/d, OR=2.88, 95% CI 0.98–8.44) (Olsen et al., 1990), California (OR= 4.7, 95% CI 1.1–21.1)
(Windham et al., 1991), and the United Kingdom (dry-cleaning operator, OR=1.63, 95% CI 1.01–2.66)
(Doyle et al., 1997), trichloroethylene in California (OR= 3.1, 95% CI 0.92–10.4) (Windham et al.,
1991), and with occupational chloroform use in Swedish laboratories (preconceptual exposure,
OR=2.3, 95% CI 0.9–5.9) (Wennborg et al., 2000).
Prenatal occupational exposure, glycol ethers, limited evidence: Employment in U.S. semiconductor
industry fabrication rooms during early pregnancy was associated with medical-record-confirmed
spontaneous abortion in a national cohort study (clinical spontaneous abortion, OR=1.43, 95% CI
0.95–2.09; clinical spontaneous abortion plus early pregnancy loss confirmed by daily urinary hCG
tests, OR=1.25, 95% CI 0.63–1.76) (Schenker et al., 1995). This study showed that spontaneous
abortion risk increased with level of exposure to glycol ethers and other photoresist and developer
chemicals (highly exposed vs. unexposed, OR=2.70, 95% CI 1.40–4.55) (Swan et al., 1995). In a
cohort study in two eastern U.S. semiconductor plants, spontaneous abortion was associated with
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 387
high exposure to ethylene glycol ether exposure during the month of conception (compared to
unexposed women, OR=2.8, 95% CI 1.4–5.6) (Correa et al., 1996). A small case-control study
nested within a cohort of UK semiconductor industry female employees revealed no association
between spontaneous abortion and prenatal work in semiconductor fabrication (OR=0.64, 95% CI
0.27–1.51); there were too few cases to assess specific solvent exposures (Elliott et al., 1999).
Prenatal occupational exposure, other or unspecified solvents, limited evidence: A meta-analysis
of 5 studies published during 1988–1992 indicated a modest association between spontaneous abor-
tion and maternal occupational exposure to organic solvents (summary OR=1.25, 95% CI 0.99–1.58)
(McMartin et al., 1998). In subsequently published studies, spontaneous abortion was associated with
prenatal occupational exposure to unspecified solvents in Finland (high vs. low 1st trimester exposure,
OR=2.3, 95% CI 1.1–4.3) (Taskinen et al., 1994) and Italy (high vs. low exposure during shoe manu-
facturing, OR=3.85, 95% CI 1.24–11.9) (Agnesi et al., 1997). Although the latter study did not assess
specific solvents, those commonly used in the local shoe industry at that time included ethylacetate,
cyclohexane, methylethylketone, and hexane. Spontaneous abortion was not associated with prenatal
occupational exposure to unspecified solvents in studies in Toronto (any solvent, crude OR=1.4, 95%
CI 0.4–4.9, calculated from data in paper) (Khattak et al., 1999) and Sweden (solvents other than
chloroform, OR=0.9, 95% CI 0.5–1.9) (Wennborg et al., 2000).
Paternal preconceptual exposure, glycol ethers, inadequate evidence: In a cohort study in two
eastern U.S. semiconductor plants, spontaneous abortion was not associated with high paternal
exposure to ethylene glycol ether exposure during the month of conception (compared to unex-
posed, OR=0.7, 95% CI 0.3–1.6) (Correa et al., 1996).
Paternal preconceptual exposure, unspecified solvents, limited evidence: A recent meta-analysis
of five epidemiologic studies published during 1984–1998 reported a statistically nonsignificant
elevated risk of spontaneous abortion related to paternal occupational solvent exposure (overall
RR=1.30, 95% CI = 0.81–2.11) (Logman et al., 2005). Among studies not included in this
meta-analysis, spontaneous abortion was associated with organic solvent exposure among men
monitored for occupational solvent exposure (OR=2.3, 95% CI 1.1–5.0) with somewhat higher
risks for painters (OR=3.3, 95% CI 1.6–6.8) and woodworkers (OR=3.8, 95% CI 1.2–12) (Taskinen
et al., 1989). In a cohort of Norwegian men in the printing industry, late spontaneous abortion (GW
20–27) was also associated with solvent exposure (with or without lead exposure, OR=5.5, 95% CI
1.8–17.2) (Kristensen et al., 1993).
Bisphenol A Bisphenol A (BPA) is a monomer used to produce polycarbonate plastics and
resins, while brominated BPA analogues are used as flame retardants.
Maternal exposure, inadequate evidence: A small case-control study reported an association
between recurrent spontaneous abortion and maternal serum bisphenol A (BPA) levels (case vs.
control mean maternal serum BPA (± SD), 2.59 ± 5.23 ng/ml vs. 0.77 ± 0.38 ng/ml, p=.02)
(Sugiura-Ogasawara et al., 2005). About half or more of 1st trimester spontaneously aborted fetuses
have chromosomal abnormalities, especially triploidy (Eiben et al., 1990), and BPA is a potent
cause of aneuploidy in mouse oocytes in vivo (Hunt et al., 2003), possibly by interfering with spindle
microtubule organization and chromosome segregation during meiosis (Can et al., 2005). In mice,
prenatal exposure to high-dose BPA produced increased fetal resorption (Morrissey et al., 1987).
Summary There was limited epidemiologic evidence for the role of environmental toxicants in
spontaneous abortion, including prenatal exposure to lead, arsenic, DDT/DDE, ethylene oxide,
active smoking, ETS, DBPs, chlorinated solvents, glycol ethers, other and unspecified solvents, and
ethylene oxide, and paternal occupational exposure to lead, chlorophenoxy herbicides other than
2,4,5-T, and other and unspecified solvents.
Stillbirths (Gestation Week ³ 20) Stillbirths are defined here at those occurring at GW 20 or
later, but some studies used other definitions (e.g., ≥28 wk). There are about 27,000 stillbirths
annually in the United States (Centers for Disease Control and Prevention, 2004). Major causes of
stillbirths include birth defects, infections, intrauterine growth restriction, gestational diabetes, and
preeclampsia (Health Canada, 2003).
Lead Maternal exposure, inadequate evidence: The Port Pirie birth cohort study found no
association between stillbirth and 2nd trimester maternal blood lead levels (only 11 cases)
388 D. T. WIGLE ET AL.
(McMichael et al., 1986). In a subsequent nested case-control study within this cohort, geometric
mean placental membrane lead levels were higher among pregnancies ending in stillbirth (2.73 μg/g,
95% CI 0.69–10.8, n=6) than controls (0.78 μg/g, 95% CI 0.61–1.00, n=22) but the difference was
not statistically significant (Baghurst et al., 1991). A large population-based case-control study found
a borderline association between stillbirth and self-reported prenatal occupational lead exposure
(Savitz et al., 1989a). A very small cohort of female survivors of childhood lead poisoning had a non-
significantly increased risk of fetal death (any gestation length), but biomarkers of lead body burden
were not measured (Hu 1991). A small case-control study in Boston reported a nonsignificant associ-
ation between stillbirth and drinking water lead levels in communities where mothers resided during
the 1st trimester (drinking water lead 2.5–80 vs. <2.5 μg/L, OR=2.1, 95% CI 0.6–7.2) (Aschengrau
et al., 1993). In a Norwegian retrospective cohort study, there was a statistically nonsignificant ele-
vated risk of late spontaneous abortion, stillbirths and neonatal deaths among women with likely
high-level lead exposure based on job title (OR=3.7, 95% CI 0.6–13) (Irgens et al., 1998).
Paternal occupational exposure, inadequate evidence: A large U.S. population-based case-con-
trol study revealed no association between stillbirth and self-reported paternal occupational lead
exposure (Savitz et al., 1989a). Among a Norwegian cohort of printers, stillbirths were associated
with potential lead exposure based on job title (OR=2.0, 95% CI =0.9–4.7) (Kristensen et al.,
1993). A cohort of male lead smelter workers in British Columbia reported statistically nonsignifi-
cant associations between stillbirth and birth defects (12 stillbirths and 30 birth defect cases were
combined for analysis) and blood lead levels of at least 25 μg/dl (Alexander et al., 1996). The
dilution of stillbirths by birth defect cases reduces the usefulness of this study. In a large Norwegian
retrospective cohort study, late spontaneous abortion, stillbirths and neonatal deaths combined
were not associated with likely paternal high-level lead exposure based on job title (OR=1.2, 95%
CI 0.7–1.9) (Irgens et al., 1998).
Mercury Maternal exposure, inadequate evidence: Stillbirths were not associated with
detectable inorganic mercury levels in drinking water in the community of prenatal residence in
Massachusetts (OR=0.7, p > .05) (Aschengrau et al., 1993) or with prenatal occupational exposure
to airborne elemental mercury in North Carolina (OR=1.42, 95% CI 0.12–37.2, calculated from
data in paper) (Elghany et al., 1997).
Inorganic arsenic Maternal exposure, drinking water, inadequate evidence: A small case-control
study in Massachusetts reported no association between stillbirth and detectable drinking water
arsenic levels in the community of prenatal residence (drinking-water arsenic >0.8 μg/L vs. unde-
tectable, OR=0.7; confidence interval not stated, p > .05; highest arsenic level was 2.6 μg/L)
(Aschengrau et al., 1993). In areas with drinking water arsenic levels above 100 μg/L, elevated
stillbirth risks were reported in ecologic studies in Hungary (rate per 1000 live births, exposed vs.
comparison region, 7.7 and 2.8, p=.03) (Borzsonyi et al., 1992) and Chile (exposed vs. comparison
region, RR= 1.72, 95% CI 1.54–1.93) (Hopenhayn-Rich et al., 2000) and in a retrospective cohort study
in Bangladesh (RR=2.24, 95% CI 0.86–6.04, calculated from data in report) (Ahmad et al., 2001).
Maternal exposure, airborne, inadequate evidence: A review of epidemiologic and toxicologic
literature noted limited evidence of associations between fetal deaths and arsenic exposure in
humans and indicated that prenatal inorganic arsenic exposure can produce fetal death in experi-
mental animals (Golub et al., 1998). Two original studies were not included in the latter review.
There was no association between stillbirth and prenatal airborne arsenic exposure (based on
self-reported work history) in a nationwide U.S. case-control study (OR=1.0, 95% CI 0.7–1.3)
(Savitz et al., 1989a). A case-control study in a Texan community with an arsenical pesticide plant
reported an association between stillbirth and estimated ambient air arsenic levels above 100 μg/m3
near the prenatal residence (OR=4.0, 95% CI 1.2–13.7) (Ihrig et al., 1998).
Paternal occupational exposure, airborne, inadequate evidence: A U.S. case-control study
found no association between stillbirth and self-reported paternal occupations with likely arsenic
exposure to airborne arsenic (OR=1.0, 95% CI 0.8–1.2) (Savitz et al., 1989a). A small case-control
study revealed an inverse association between stillbirth and parental employment in a Swedish
copper smelter (no data reported on maternal/paternal status but high-exposure jobs were held
mainly by men) (Wulff et al., 1995).
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 389
Cadmium Maternal exposure, inadequate evidence: In a Massachusetts case-control, still-
births were not associated with low-level drinking water cadmium levels in communities of pre-
natal residence (≥0.4 vs. <0.4 μg/L, OR=1.2, p > .05) (Aschengrau et al., 1993). A Swedish
ecologic study demonstrated no overall increased risk of stillbirths among women living in munic-
ipalities with soil cadmium levels exceeding 1.6 μg/g (OR= 0.97, 95% CI 0.72–1.29); in the
municipality with the highest soil cadmium level (12 μg/g), stillbirth risk was elevated (OR=2.17,
95% CI 1.06–4.49) (Landgren, 1996).
PCBs Maternal exposure, high-level exposure, inadequate evidence: A retrospective cohort
study of women exposed to high levels of PCBs, PCDFs and related toxicants during the Yucheng
incident revealed no increased risk of stillbirths (maternal serum PCB >46 vs. ≤46 ng/L, OR=1.35,
95% CI 0.35–5.26) (Yu et al., 2000).
Maternal exposure, low-level exposure, inadequate evidence: Reviewers found insufficient
evidence for an association between stillbirth and background PCB exposure (Longnecker et al.,
1997). In a Swedish retrospective cohort study of fishing families, stillbirth was not associated with
residence in a region with fish contaminated by relatively high PCB concentrations (OR=1.58, 95%
CI 0.50–5.04) (Axmon et al., 2000).
TCDD Paternal occupational exposure, inadequate evidence: Compared to Vietnam veterans
not exposed to Agent Orange, there were statistically nonsignificant elevated risks of stillbirth (odds
ratios calculated from data in paper) among exposure groups categorized as background (current and
baseline TCDD ≤10 ng/L, OR= 1.89, 95% CI 0.68–5.12) or low (current TCDD ≤ 10 ng/L and base-
line ≤110 ng/L, OR= 1.90, 95% CI 0.64–5.43); the odds ratio for the high TCDD category (current
>10 ng/L and baseline >110 ng/L) was not calculated as there was only 1 exposed case father (Wolfe
et al., 1995). Among a cohort of British Columbia sawmill workers, stillbirths were not associated with
paternal occupational exposure to chlorophenate wood preservatives contaminated by TCDD and
related toxicants (per 100-h increment of exposure up to 3 mo before conception, OR=1.0, 95% CI
0.97–1.063) (Dimich-Ward et al., 1996).
Pesticides Maternal exposure, nonchlorophenoxy herbicides, inadequate evidence: In a California
case-cohort study, there was an association of borderline statistical significance between stillbirths/
neonatal deaths from birth defects and agricultural use of paraquat during GW 3–8 (OR= 1.8, 95%
CI 0.9–3.9) in the same square mile section as the prenatal residence (Bell et al., 2001b).
Maternal exposure, DDT/DDE, limited evidence: A review of available studies found limited
evidence of an association between fetal deaths and biomarkers of maternal DDT/DDE exposure
(Weselak et al., 2006). Among recent studies included in their review, a retrospective cohort study
based on the U.S. Collaborative Perinatal Project reported a nonmonotonic dose-response relationship
between fetal deaths of any gestation length and prenatal serum DDE levels (per 60 μg/L maternal
serum DDE increment, OR=1.4, 95% CI 1.1–1.6) (Longnecker et al., 2005). The latter authors
stated that the association with maternal serum DDE was similar for spontaneous abortions and still-
births but did not include supporting data. A subsequently reported Australian retrospective cohort
study observed no association between fetal death (any gestation length) and breast milk DDT or
DDE levels (e.g., 3rd vs. 1st tertile DDE, OR=0.76, 95% CI 0.41–1.39) (Khanjani & Sim, 2006).
Maternal exposure, organophosphate insecticides, inadequate evidence: A nested case-control
study reported an elevated stillbirth risk among women living in regions sprayed with malathion 1–4 wk
before the outcome (OR=1.95, 95% CI 0.88–4.35) (Thomas et al., 1992). A California-wide study
revealed that agricultural organophosphate application (mainly insecticides) during GW 3–8 in the
same square mile section as the prenatal residence was associated with stillbirths/neonatal deaths
from birth defects (OR=2.9, 95% CI 1.3–6.4) (Bell et al., 2001b) but not with other stillbirths (1st
trimester exposure, OR=1.0, 95% CI 0.7–1.3) (Bell et al., 2001c). Although suggestive, these find-
ings require confirmation and exploration of dose-response relationships.
Maternal exposure, pyrethroid insecticides, inadequate evidence: Stillbirths/neonatal deaths
from causes other than birth defects in California were not associated with agricultural pyrethroid
insecticide application in the same square mile section as the maternal residence (1st trimester use,
OR=1.0, 95% CI 0.6–1.6) (Bell et al., 2001c). However, there was an association between still-
births/neonatal deaths from birth defects and agricultural pyrethroid insecticide application during
390 D. T. WIGLE ET AL.
GW 3–8 in the same square mile section as the maternal residence (OR=4.9, 95% CI 1.9–12.9);
this analysis excluded subjects for whom other pesticide classes were also used during GW 3–8 (Bell
et al., 2001a, 2001b). The California study used a highly detailed statewide pesticide use database
and did not have to rely on individual pesticide exposure recall; however, it did not assess potential
relationships between stillbirths/neonatal deaths and preconceptual paternal exposure.
Maternal exposure, organochlorine insecticides, inadequate evidence: An Australian retrospec-
tive cohort study found no association between fetal death (any gestation length) and breast milk
dieldrin (3rd vs. 1st tertile, OR=0.73, 95% CI 0.42–1.27), heptachlor epoxide (OR= 0.82, 95% CI
0.48–1.39) or oxychlordane levels (OR=0.54, 95% CI 0.29–1.00) (Khanjani & Sim, 2006).
Maternal exposure, unspecified insecticides, inadequate evidence: A case-control study in
California agricultural counties revealed an association of borderline statistical significance between
all stillbirths/neonatal deaths and 1st trimester indoor use of insecticides at the maternal residence
(OR=1.4, 95% CI 0.9–2.3) but not insecticide use on pets (OR=1.0, 95% CI 0.5–1.9) (Pastore
et al., 1997). There was a statistically nonsignificant elevated risk of stillbirths among Sudanese
women who reported prenatal spraying of pesticides (mainly insecticides) using hand pumps
indoors at home (OR=1.6, 95% CI 0.8–3.3) (Taha & Gray, 1993).
Maternal exposure, fungicides, inadequate evidence: An Australian retrospective cohort study
found no association between fetal death (any gestation length) and breast HCB levels (3rd vs. 1st
tertile, OR=0.49, 95% CI 0.26–0.90) (Khanjani & Sim, 2006).
Maternal exposure, unspecified pesticides, inadequate evidence: Reviewers found limited epi-
demiologic evidence for associations between fetal deaths and maternal exposure to unspecified
pesticides (Arbuckle & Sever, 1998). Among studies of stillbirths published before these reviews,
prenatal exposures linked to elevated stillbirth risks included maternal employment in agriculture
or horticulture (compared to all women in cohort, RR=5.55, 95% CI 1.51–14.2, based on 4
exposed case mothers) (McDonald et al., 1988), pesticide use at work (OR=1.6, 95% CI 1.3–2.1)
or home (OR=1.5, 95% CI 1.3–1.7) (Savitz et al., 1989b), occupational pesticide use in China
(lower 95% confidence limit = 4.48, p < .05; OR and upper 95% confidence limit not available
as there were 7 exposed case and 0 exposed control mothers) (Zhang et al., 1992), prenatal spraying
of pesticides on Sudanese farms (OR= 3.6, 95% CI 1.6–8.0) (Taha & Gray, 1993), and maternal
occupational pesticide use during the 1st trimester (OR=2.7, 95% CI 1.5–4.8) or 2nd trimester
(OR=2.2, 95% CI 1.0–4.9) (Pastore et al., 1997). In the latter study, there was a relatively strong
association between the subgroup of stillbirths and neonatal deaths from complications of placenta
or cord and 1st trimester maternal occupational pesticide use (OR=4.8, 95% CI 2.0–11.4).
A recent Danish pregnancy cohort study revealed a statistically nonsignificant elevated risk of fetal
loss at any gestation length related to prenatal occupation as gardeners (OR=1.7, 95% CI 0.7–4.0, 5
exposed case mothers) (Zhu et al., 2006). Studies reporting no association between stillbirths and mater-
nal pesticide exposure indices included case-control studies in the United States (occupation in agricul-
ture, forestry or fishing, OR=0.8, 95% CI 0.5–1.3) (Savitz et al., 1989a), Colombia (employment in
semi-enclosed horticultural greenhouses, OR=0.99, 95% CI 0.66–1.48) (Restrepo et al., 1990b), and
California (1st trimester maternal residential proximity to agricultural crops, OR=1.0, 95% CI 0.9–
1.2; 1st trimester garden pesticide use, OR = 0.7, 95% CI 0.6–3.1) (Pastore et al., 1997). Although
suggestive, the heterogeneity of exposure indices precludes firm conclusions.
Paternal occupational exposure, 2,4,5-T and chlorophenate wood preservatives, inadequate
evidence: See discussion of paternal occupational TCDD exposure earlier.
Paternal occupational exposure, unspecified pesticides, inadequate evidence: A retrospective
cohort study in Colombia reported no association between stillbirths and paternal employment in
semi-enclosed floriculture greenhouses in which a total of 127 different types of pesticides were used
(OR=0.87, 95% CI 0.42–1.83) (Restrepo et al., 1990b). Partners of male pesticide applicators had a
substantially increased risk of stillbirths in an Indian retrospective cohort study (crude OR=4.97, 95%
CI 3.85–6.42, calculated from data in paper); exposed men used backpacks to apply organochlorine
and other pesticides on cotton (Rupa et al., 1991). There may have been recall bias in this study as
only 2 birth defects were reported among 3016 pregnancies in the unexposed group compared to
128 among 4240 pregnancies in the exposed group. In a retrospective cohort of Norwegian farm
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 391
families, stillbirth risk (based on official records) was not elevated (compared to nonfarm families,
RR=0.88, 95% CI 0.79–0.98) (Kristensen et al., 1997a). In a Spanish retrospective cohort study of
pregnancies conceived during the pesticide-use season (April–September), stillbirth (based on official
records) was associated with paternal occupation in agriculture (stillbirths caused by birth defects,
OR=1.62, 95% CI 1.01–2.60; other stillbirths, OR=1.35, 95% CI 1.11–1.65) (Regidor et al., 2004).
Although suggestive, the heterogeneity of exposure indices precludes firm conclusions.
Tobacco smoke Maternal active smoking, limited evidence: The U.S. Surgeon General
concluded that there is suggestive evidence of a causal relationship between stillbirth and maternal
active smoking (U.S. Department of Health and Human Services, 2001).
Maternal exposure, ETS, inadequate evidence: Stillbirths were not associated with ETS expo-
sure at home (OR=1.1, 95% CI 0.6–2.4) or at work (OR = 1.2, 95% CI 0.6–2.4) among Swedish
women (Ahlborg & Bodin, 1991). A pregnancy cohort study in California reported a statistically
nonsignificant association between stillbirth (defined as GW ≥20) and maternal serum cotinine
levels during early pregnancy (5th vs. 1st quintile, OR=3.36, 95% CI 0.81–14.0) (Kharrazi et al.,
2004). An expert panel review found sparse data on stillbirths and inadequate evidence of an asso-
ciation with prenatal ETS exposure (California Environmental Protection Agency, 2005).
Outdoor air pollution Maternal exposure, inadequate evidence: Daily stillbirths were weakly
associated with NO2 levels in ecologic studies in Brazil (5-pollutant model, daily stillbirths vs. NO2
(mg/m3), regression coefficient=0.0012 ± 0.0004 (Pereira et al., 1998) and the Czech Republic
(per NO2 increment of 50 μg/m3, OR=1.21, 95% CI 0.89–1.64) (Bobak & Leon, 1999). Stillbirths
in the United Kingdom were not associated with prenatal residence proximity to coke works (major
point sources of particulate air pollution) (≤2 vs. >2 km, OR=0.94, 95% CI 0.78–1.12) (Dolk et al.,
2000). A recent review concluded that the few epidemiologic studies of stillbirth provided inade-
quate evidence for an association with ambient air pollution (Glinianaia et al., 2004a).
Drinking water disinfection by-products Maternal exposure, limited evidence: Reviewers
found limited and fairly consistent evidence for an association between stillbirth and prenatal DBP
exposure (Bove et al., 2002; Graves et al., 2001; Nieuwenhuijsen et al., 2000). Studies in Nova
Scotia reported nonmonotonic dose-response relationships between stillbirths and drinking water
THM levels (≥100 vs. <50 μg/L, OR=1.66, 95% CI 1.09–2.52) (Dodds et al., 1999), especially
among the subset of stillbirths caused by asphyxia arising mainly from abruptio placenta (≥100 vs.
<50 μg/L, OR=4.57, 95% CI 1.93–10.8) (King et al., 2000). The association between stillbirth and
THMs was mainly related to BDCM (≥20 vs. <5 μg/L, OR=1.98, 95% CI 1.23–3.49) (King et al., 2000).
An expanded study in Nova Scotia and eastern Ontario revealed that stillbirths were associated with tap
water total THM levels (≥80 vs. <1 μg/L, OR=2.2, 95% CI 1.1–4.4) and with total THM exposure from
all sources including showering/bathing (5th vs. 1st quintile, OR= 2.4, 95% CI 1.2–4.6) (Dodds et al.,
2004). Although this study observed nonmonotonic dose-response relationships, the authors noted that
the highest exposure subgroups consistently had the highest stillbirth risks. A retrospective cohort study in
England reported slightly increased stillbirth risks among women living in regions with the highest tap
water THM levels (≥60 vs. <30 μg/L, OR=1.11, 95% CI 1.00–1.23) (Toledano et al., 2005).
Drinking water nitrate Maternal exposure, inadequate evidence: A nested case-control study in
Massachusetts revealed no association between stillbirths and drinking water nitrate levels in the commu-
nity of maternal residence at birth (0.3–4.5 vs. <0.2 mg/L, OR=0.8, p > .05) (Aschengrau et al., 1993).
Hazardous waste disposal sites Maternal exposure, inadequate evidence: A case-control study
within the 1988 U.S. National Maternal and Infant Health Survey reported a weak association
between stillbirth and prenatal residential proximity to U.S. EPA National Priority List2 (NPL)
hazardous waste disposal sites (maternal residence ≤1.6 km from a NPL site, OR= 1.14, 95% CI
0.95–1.36) (Sosniak et al., 1994). In record-based retrospective cohort studies in the United
Kingdom, stillbirths were not associated with prenatal residential proximity to hazardous industries
(per unit change in an inverse distance squared function, OR=0.95, 95% CI 0.87–1.00) (Dummer
et al., 2003b), incinerators (per unit change in an inverse distance squared function, OR=1.04,
95% CI 0.90–1.19) (Dummer et al., 2003a), or hazardous waste landfill sites (<2 km from sites in
2The EPA’s list of the most serious uncontrolled or abandoned hazardous waste disposal sites in the United States.
392 D. T. WIGLE ET AL.
the United Kingdom, OR=0.99, 95% CI 0.95–1.03; <2 km from landfill sites in Scotland,
OR=0.99, 95% CI 0.95–1.03 [Morris et al., 2003]).
Solvents Maternal occupational exposure, various and unspecified solvents, inadequate
evidence: A large population-based U.S. case-control study reported weak associations between
stillbirth and likely prenatal occupational exposure (inferred from job titles) to benzene (OR=1.3,
95% CI 1.0–1.8) and petroleum (OR=1.4, 95% CI 1.0–1.9) but not alcohols/glycols or chlorinated
hydrocarbons (Savitz et al., 1989a). A cohort study in Montreal revealed an association between
stillbirth and maternal 1st trimester occupational solvent exposure (OR=2.76, p < .01) (McDonald
et al., 1988). A case-control study in Brazil observed no association between stillbirth and prenatal
residential proximity to a petrochemical plant (<10 vs. ≥30 km, OR=0.71, 95% CI 0.20–2.56)
(keOliveira et al., 2002).
Paternal occupational exposure, various and unspecified solvents, inadequate evidence: A large
population-based U.S. case-control study observed no association between stillbirth and likely
paternal occupational exposure (inferred from job titles) to benzene, petroleum, alcohols/glycols or
chlorinated hydrocarbons (Savitz et al., 1989a). Fetal deaths (mainly stillbirths) were not associated
with paternal occupational solvent exposure inferred from job titles recorded on birth records in
Washington State (painting, OR=0.9, 95% CI 0.8–1.1; autobody work, OR=1.0, 95% CI 0.8–1.2;
printers, OR=1.1, 95% CI 0.8–1.3) (Daniell & Vaughan, 1988). A Norwegian cohort study of male
printers showed no association between stillbirth and occupational solvent exposure (OR=0.9,
95% CI 0.4–2.2) (Kristensen et al., 1993).
Summary There was limited epidemiologic evidence for the role of environmental toxicants in
stillbirths including prenatal exposure to DDT/DDE or DBPs.
Preterm Birth Preterm birth, defined as gestation length <37 wk, accounts for 75–85% of all
perinatal mortality in Canada (Public Health Agency of Canada, 2005). Affected infants have
increased risks of neurodevelopmental handicaps, infections, chronic respiratory disease, and
ophthalmologic problems (Health Canada, 2003). Preterm birth differs from intrauterine growth
restriction (see next section) with regard to etiology and outcome.
Lead Maternal exposure, limited evidence: Reviewers found limited and somewhat inconsis-
tent epidemiologic evidence for an association between preterm birth and prenatal lead exposure
(Andrews et al., 1994). Preterm birth was generally associated with maternal but not paternal blood
lead levels (Figure 1). In the Port Pirie birth cohort study, there was a monotonic dose-response
relationship between preterm birth and prenatal blood lead levels with a 4.4-fold increased risk in
the highest exposure category (maternal blood lead ≥14 vs. <8 μg/dl, OR=4.4, 95% CI 1.2–17)
(McMichael et al., 1986). A nested case-control study within this cohort found somewhat higher
placental membrane lead levels among pregnancies ending in preterm birth (geometric mean lead
concentration, cases vs. controls, 1.24 μg/g [95% CI 0.91–1.67] vs. 0.78 μg/g [95% CI 0.61–1.00];
Baghurst et al., 1991). In two other birth cohort studies, preterm birth was not associated with cord
blood lead in Boston (per 1 μg/dl increment, OR=0.98, 95% CI 0.93–1.02) (Bellinger et al., 1991)
or with placental, prenatal blood, or cord blood lead levels in the former Yugoslavia (per 2nd
trimester maternal blood increment (μmol/L), OR=0.99, 95% CI 0.97–1.01) (Factor-Litvak et al.,
1991). All of these studies adjusted for prenatal smoking and other potential confounders.
In a Norwegian retrospective cohort study, preterm birth was related to likely high-level
occupational lead exposure based on job title (OR=1.9, 95% CI 1.1–3.3) (Irgens et al., 1998).
A case-cohort study in Mexico City reported a nonmonotonic dose-response relationship between
preterm birth and cord blood lead levels but only among primiparous women (cord blood lead ≥15
vs.<5.1 μg/dl, OR= 2.6, 95% CI 1.0–6.7) (Torres-Sanchez et al., 1999). Gestation length was
inversely associated with placental lead concentration in a small Spanish case control study with
limited statistical analysis (Pearson’s r = –.32, p= .002) (Falcon et al., 2003). A retrospective cohort
study of occupationally exposed persons in Taiwan revealed a dose-response relationship between
preterm birth and maternal prenatal blood lead levels (maternal blood lead 10–19 vs. <10 μg/dl,
RR=1.97, 95% CI 0.92–3.86; blood lead ≥20 vs. <10 μg/dl, RR =1.86, 95% CI 0.68–4.28;
p-trend = .06) (Chen et al., 2006). In a Californian retrospective cohort study, women with maxi-
mum pregnancy blood lead levels of at least 10 μg/dl had a substantially increased risk of preterm
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 393
birth (OR=3.2, 95% CI 1.2–7.4, compared to women with lower maximum levels); among women
with blood lead levels of at least 10 μg/dl, gestation length decreased by an average of 1 day per
increment of 1 μg/dl in 2nd trimester maximum maternal blood lead level (Jelliffe-Pawlowski et al.,
2006). Maternal exposures to airborne lead emissions in Shoshone County, Idaho (during a 15-mo
period when air emissions were high because of a damaged bag house), was not associated with
increased risk of preterm birth (OR=0.68, 90% CI 0.34–1.35) (Berkowitz et al., 2006).
Paternal occupational exposure, inadequate evidence: Three studies that inferred paternal lead
exposure from job histories yielded little or no evidence of an association with preterm birth (Irgens
et al., 1998; Kristensen et al., 1993; Savitz et al., 1989a). In a U.S. retrospective cohort study of
occupationally exposed men, there was a moderately strong association between preterm birth and
blood lead levels of at least 25 μg/dl for at least 5 yr (OR=3.0, 95% CI 1.4–6.8) (Lin et al., 1998).
A retrospective cohort study of persons occupationally exposed to lead in Taiwan reported no asso-
ciation between preterm birth and preconceptual blood lead levels of at least 20 μg/dl (OR=0.6,
95% CI 0.2–1.3) (Chen et al., 2006).
Inorganic arsenic Maternal occupational exposure, airborne arsenic, inadequate evidence:
A nation-wide case-control study in the United States found no association between preterm birth and
prenatal occupations likely exposed to airborne arsenic (OR=0.7, 95% CI 0.4–1.4) (Savitz et al., 1989a).
Maternal exposure, drinking water, inadequate evidence: Preterm births were associated with
prenatal residence in a region with well-water arsenic levels above 100 μg/L in a retrospective
cohort study in Bangladesh (RR= 2.65, 95% CI 1.10–6.58, calculated from data in report) (Ahmad et al.,
2001). In an ecologic study, preterm birth was not associated with prenatal residence in regions of
Taiwan with high well water arsenic levels (OR=1.10, 95% CI 0.91–1.33) (Yang et al., 2003b).
FIGURE 1. Preterm birth vs. parental lead exposure (m=prenatal, p=paternal, Pb =lead, BPb=blood lead, occup = occupational).
Study
Huel
McMichael
Savitz
Bellinger
Factor
Irgens
Torres
Chen
Jelliffe
Berkowitz
Savitz
Irgens
Lin
Chen
Grp
m
m
m
m
m
m
m
m
m
m
m
m
m
m
p
p
p
p
p
Compare
hair Pb > 14
BPb 08–10
BPb 11–13
BPb 14+
occup
BPb 15+ vs. <5
per 20 ug/dL
occup L
occup M
BPb 15+ vs. <5
10–19 ug/dL
BPb 20+ vs. <10
BPb 10+ vs. <10
airborne Pb
occup
occup H
occup L
25+ ug/dL
BPb 20+ vs. <10
Yr
1981
1986
1989
1991
1991
1998
1999
2006
2006
2006
1989
1998
1998
2006
Odds ratio and 95% CI
0.1 0.2 0.5 12 5
10
394 D. T. WIGLE ET AL.
Paternal occupational exposure, inadequate evidence: A U.S. nation-wide case-control study
observed no association between preterm birth and paternal occupations likely exposed to airborne
arsenic (OR=1.1, 95% CI 0.7–1.7) (Savitz et al., 1989a).
Cadmium Maternal exposure, inadequate evidence: A small retrospective cohort study in France
revealed a statistically nonsignificant elevated risk of preterm birth related to maternal hair cadmium
(≥0.42 vs. <0.42 μg/g, OR=2.69, 95% CI 0.53–15.0, calculated from data in paper) (Huel et al., 1981).
A birth cohort study in a lead smelter town in the former Yugoslavia reported no association between
gestation length and placental cadmium among nonsmoking women (change in gestation length per pla-
centa cadmium increment of 1 nmol/g, 4.30 d, 95% CI –4.9 to 13.5) (Loiacono et al., 1992). A Swedish
ecologic study demonstrated no association between preterm birth and maternal residence in munici-
palities with soil cadmium levels exceeding 1.6 μg/g (OR=0.93, 95% CI 0.84–1.03); the risk in the
municipality with the highest soil cadmium level (12 μg/g) was not elevated (OR=0.86, 95% CI 0.61–
1.21) (Landgren, 1996). A small birth cohort study in a cadmium-polluted region in Japan showed an
association between preterm birth and prenatal urinary cadmium levels (≥2 vs. <2 μg/g creatinine,
crude OR=7.32, 95% CI 1.27–45.5, calculated from data in paper) (Nishijo et al., 2002). A small birth
cohort study in a cadmium-polluted region of China reported no association between preterm birth
among nonsmoking women and maternal or cord blood cadmium levels (cord blood cadmium >0.40
vs. ≤0.40 μg/L, OR=1.46, 95% CI 0.23–9.56, calculated from data in paper) (Zhang et al., 2004).
PCBs Maternal occupational exposure, limited evidence: Among women prenatally exposed
to airborne PCBs during capacitor production, there was an inverse dose-response relationship of bor-
derline statistical significance between gestation length and estimated serum PCB levels (β=–1.1 d,
90% CI –2.0 to –0.1) (Taylor et al., 1989).
Maternal high-level environmental exposure, inadequate evidence: Follow-up to 1993–1994 of
women exposed to high-levels of PCBs during the 1979 Yucheng incident revealed an elevated
prevalence of stillbirths (exposed vs. unexposed, 4.2 vs. 1.7%, p= .07) (Yu et al., 2000). This finding
is consistent with an elevated risk possibly diluted by declining body burden of PCBs, PCDFs and
related toxicants over 15 yr.
Maternal low-level environmental exposure, inadequate evidence: After adjustment for the
relative concentration of docosahexaenoic acid (an n–3 polyunsaturated fatty acid in seafood) in
cord serum phospholipids, gestation length was not associated with prenatal serum PCB levels in a
small Faroe Islands birth cohort study (Grandjean et al., 2001). A small Spanish birth cohort study
revealed no association between preterm birth and mean maternal serum PCB levels (Ribas-Fito et al.,
2002). Preterm birth was weakly associated with maternal residence in zip code areas of New York State
with PCB-contaminated hazardous waste disposal sites (crude OR=1.10, 95% CI 1.08–1.11); this study
did not use exposure biomarkers and did not adjust for potential confounders (Baibergenova et al.,
2003). In a California pregnancy cohort study conducted during the 1960s (when population serum PCB
levels were substantially higher than currently) gestation length was inversely associated with prenatal
serum PCB (per natural log serum PCB increment, β=–3.9 ± 2.0 d) (Hertz-Picciotto et al., 2005). How-
ever, in a similar study with mothers recruited in 12 U.S. cities during 1959–1965, preterm birth was not
associated with prenatal serum PCB levels (≥4 vs. <2 μg/L, OR=1.11, 95% CI 0.55–2.24) (Longnecker
et al., 2005). There was also no association in cohort studies of Great Lakes fish eaters (change in gesta-
tion length per 2.7-fold maternal serum PCB increment, β=–0.08 wk [95% CI –0.75 to 0.59];
Weisskopf et al., 2005) or a representative sample of births in Victoria, Australia (preterm birth, breast
milk PCB 10–49 vs. <10 μg/kg lipid, OR=1.41, 95% CI 0.25–7.96; ≥50 vs. <10 μg/kg lipid, OR=2.30,
95% CI 0.40–13.3; p-trend=.43) (Khanjani & Sim, 2007).
TCDD Maternal exposure, inadequate evidence: Preterm birth during an 8-year follow-up of
women exposed at Seveso was not associated with maternal serum TCDD levels (per log incre-
ment, OR=1.3, 95% CI 0.7–2.3); there was also no association between gestation length and
maternal serum TCDD (per log increment, β=–1.2 d, 95% CI –2.9 to 0.5) (Eskenazi et al., 2003).
Paternal occupational exposure, inadequate evidence: Preterm birth was not associated with
exposure to potentially TCDD-contaminated chlorophenate wood preservatives among male
sawmill workers (per 100-h increment in cumulative exposure up to 3 mo before conception,
OR=1.00, 95% CI 0.99–1.001) (Dimich-Ward et al., 1996) or with paternal serum TCDD levels in
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 395
the study of U.S. veterans exposed to Agent Orange (serum TCDD at conception ≥79 vs. ≤10 pg/g
lipid, OR=1.36, 95% CI 0.75–2.39) (Michalek et al., 1998). Similarly, preterm birth was
not related to serum TCDD levels among men exposed during production of trichlorophenol and
derivatives such as 2,4,5-T (per log serum TCDD increment, OR = 0.8, 95% CI 0.6–1.1) (Lawson
et al., 2004).
Pesticides Maternal exposure, chlorophenoxy herbicides, inadequate evidence: A large
ecologic study in four U.S. Midwest states revealed no relationship between preterm birth and
maternal residence in high-wheat rural counties (a proxy for chlorophenoxy herbicide exposure)
(standardized incidence ratio [SIR]=1.05, 95% CI 0.95–1.16) (Schreinemachers, 2003).
Maternal exposure, DDT/DDE, limited evidence: A large U.S. retrospective cohort study
reported a monotonic dose-response relationship between preterm birth and prenatal serum DDE
levels (≥60 vs. <15 μg/L, OR= 3.1, 95% CI 1.8–5.4; p-trend=.0001); this study exploited
preserved prenatal blood samples collected during 1959–1965 when population DDT exposure
was much higher than now (Longnecker et al., 2001). A small birth cohort study in Spain revealed
an association between preterm birth and cord serum DDE levels (mean cord serum DDE, cases vs.
controls, 2.40 vs. 0.80 μg/L, p < 0.05) (Ribas-Fito et al., 2002). A case-cohort study in Mexico City
reported a nonmonotonic dose-response relationship between preterm birth and 1st trimester
maternal serum DDE levels (3rd vs. 1st tertile, OR= 1.7, 95% CI 0.8–3.3, p-trend= 0.17) (Torres-
Arreola et al., 2003). A retrospective cohort study of women who consumed Great Lakes fish
showed no relationship between gestation length and maternal serum DDE levels (change in gesta-
tion length per natural log serum DDE increment, β=0.03 wk, 95% CI –0.50 to 0.57) (Weisskopf
et al., 2005). In an Australian retrospective cohort study, preterm birth was not associated with
breast milk DDT or DDE levels (2nd vs. 1st tertile DDE, OR= 0.85, 95% CI 0.39–1.84; 3rd vs. 1st
tertile, OR=1.03, 95% CI 0.46–2.29) (Khanjani & Sim, 2006). A birth cohort study in the Salinas
Valley of California revealed no association between gestation length and maternal serum DDE (per
log10 serum DDE increment, β=–0.10 wk, 95% CI –0.40 to 0.20, p=.51) (Fenster et al., 2006).
Although 3 recent cohort studies found no relationship, the steep decline of population serum or
breast milk DDT/DDE levels over the past 40 yr may have contributed to negative findings.
Maternal exposure, other organochlorine insecticides, inadequate evidence: A case-cohort
study in Mexico City found a borderline dose-response relationship between preterm birth and 1st
trimester maternal serum β-hexachlorocyclohexane (β-HCH) levels (3rd vs. 1st tertile, OR= 1.9, 95% CI
0.9–3.7, p-trend=.08) (Torres-Arreola et al., 2003). A California study reported an inverse dose-response
relationship between gestation length and maternal serum lipid-adjusted HCB levels (change in gesta-
tion length per natural log serum HCB increment, β=–0.47 wk, 95% CI –0.95 to –0.002, p=.05)
(Fenster et al., 2006). In an Australian retrospective cohort study, preterm birth was not associated
with breast milk dieldrin (3rd vs. 1st tertile, OR=1.22, 95% CI 0.61–2.45), heptachlor epoxide
(OR=1.02, 95% CI 0.49–2.11), oxychlordane (OR = 0.93, 95% CI 0.39–2.22) or HCB (OR= 1.27,
95% CI 0.54–3.00) (Khanjani & Sim, 2006).
Maternal exposure, organophosphate insecticides, inadequate evidence: A pregnancy cohort
study in New York found no association between gestation length and mean 3rd trimester maternal
urinary metabolite levels of chlorpyrifos (Berkowitz et al., 2004).
Maternal exposure, other specified insecticides or repellents, inadequate evidence: Preterm
birth was not associated with prenatal DEET exposure in a small Thai randomized trial (OR=1.00,
95% CI 0.54–1.85) (McGready et al., 2001). A pregnancy cohort study in New York found no asso-
ciation between gestation length and mean 3rd trimester maternal urinary pyrethroid metabolite
levels (Berkowitz et al., 2004).
Maternal exposure, fungicides, inadequate evidence: A case-cohort study in Mexico City found
no association between preterm birth and 1st trimester maternal serum HCB levels (OR=0.9, 95%
CI 0.5–1.8, p-trend=.80) (Torres-Arreola et al., 2003). A birth cohort study in the Salinas Valley of
California revealed an inverse association between gestation length and maternal serum HCB (per
log10 serum HCB increment, β=–0.47 wk, 95% CI –0.95 to –0.002, p=.05) (Fenster et al., 2006).
A pregnancy cohort study in New York found no association between gestation length and mean
3rd trimester maternal urinary pentachlorophenol levels (Berkowitz et al., 2004).
396 D. T. WIGLE ET AL.
Maternal exposure, unspecified pesticides, inadequate evidence: Preterm birth was not associated
with self-reported prenatal pesticide exposure at home (OR=1.0, 95% CI 0.7–1.4) or work (OR=1.1,
95% CI 0.6–2.1) in a large population-based case-control study (Savitz et al., 1989b). A nested case-con-
trol study reported an association between preterm birth and prenatal occupational pesticide exposure
in Colombian greenhouses (OR=1.86, 95% CI 1.59–2.17) (Restrepo et al., 1990b). A small cohort study
of mainly Hispanic pregnant women in an intense agricultural region of California revealed a lower than
expected preterm birth rate (5.6 vs. 8.9%) (Willis et al., 1993). A Danish birth cohort study showed no
association between preterm birth and prenatal occupation as gardeners (OR=1.4, 95% CI 0.8–2.4) or
farmers (OR=1.0, 95% CI 0.5–1.8) or with direct contact with pesticides at work (OR=0.7, 95% CI
0.1–5.7) (Zhu et al., 2006). The heterogeneity of exposure indices precludes firm conclusions.
Paternal occupational exposure, 2,4,5-T or chlorophenate wood preservatives, inadequate
evidence: See discussion of paternal occupational TCDD exposure earlier.
Paternal occupational exposure, other chlorophenoxy herbicides, inadequate evidence: In a
retrospective cohort study of Ontario farm families, preterm birth was associated with preconcep-
tual paternal yard use (OR=2.5, 95% CI 0.9–7.3) but not crop use (OR=1.4, 95% CI 0.5–3.6) of
chlorophenoxy herbicides (Savitz et al., 1997a).
Paternal occupational exposure, nonchlorophenoxy herbicides, inadequate evidence: In a ret-
rospective cohort study of Ontario farm families, preterm birth was associated with preconceptual
paternal atrazine use in yards (OR=4.9, 95% CI 1.6–15.0); there were also statistically nonsignifi-
cant elevated risks of preterm birth related to use on crops of atrazine (OR=2.4, 95% CI 0.8–7.0)
or glyphosate (OR=2.4, 95% CI 0.8–7.9) (Savitz et al., 1997a). As in most other studies, pesticide-
specific risk estimates were not adjusted for exposure to other specified pesticides.
Paternal occupational exposure, unspecified pesticides, inadequate evidence: Preterm birth
was not associated with self-reported paternal pesticide exposure in the home (OR = 1.0, 95%
CI 0.7–1.4), outdoors at home (OR=0.9, 95% CI 0.4–1.9) or at work (OR=1.1, 95% CI 0.7–1.8) in
a large U.S. population-based case-control study (Savitz et al., 1989b) or with agricultural work in a
large case-control study in North Carolina (OR= 0.5, 95% CI 0.2–1.4) (Savitz et al., 1997b). In a
nested case-control study, preterm birth was associated with paternal occupational pesticide use in
Colombian greenhouses (OR=2.75, 95% CI 2.01–3.76) (Restrepo et al., 1990b). A large Norwegian
retrospective cohort study found no association between preterm birth and parental employment in
farming (compared to nonfarm families, OR=0.95, 95% CI 0.91–0.99) (Kristensen et al., 1997a).
Tobacco smoke Maternal active smoking, sufficient evidence: The U.S. Surgeon General
concluded that there is suggestive evidence of a causal relationship between maternal active
smoking and preterm delivery and shortened gestation (U.S. Department of Health and Human
Services, 2004).
Maternal ETS exposure, sufficient evidence: Reviewers noted limited evidence of an association
between preterm birth and prenatal ETS exposure (Lindbohm et al., 2002) but an expert panel
recently concluded that there is sufficient evidence (California Environmental Protection Agency,
2005). Among reviewed studies, a Finnish report indicated a dose-response relationship between
preterm birth and maternal hair nicotine levels in segments corresponding to 3rd trimester exposure
(≥4.0 vs. <0.75 μg/g, OR=6.12, 95% CI 1.31–28.7; per μg/g (hair nicotine analyzed as continuous
variable), OR=1.22, 95% CI 1.07–1.39) (Jaakkola et al., 2001a), and a large cohort study of
nonsmoking Californian women noted a nonmonotonic dose-response relationship with 2nd tri-
mester maternal serum cotinine levels (5th vs. 1st quintile, OR= 1.78, 95% CI 1.01–3.13) (Kharrazi
et al., 2004). In a South African cohort study, preterm birth among nonsmoking women was not
associated with number of smokers in home (mean gestation lengths among unexposed women
and those exposed to 1 or 2+ smokers at home, respectively, were 38.4, 38.2, and 38.2 wk) (Steyn
et al., 2006). The U.S. Surgeon General reviewed eight available studies and concluded that there
was suggestive evidence of a causal relationship between prenatal ETS exposure and preterm birth
(U.S. Department of Health and Human Services, 2006).
Outdoor air pollution Maternal exposure, major ambient pollutants, limited evidence:
Reviewers noted limited evidence from studies published up to 2001 for a weak association
between preterm birth and prenatal exposure to ambient air pollutants including particulate matter
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 397
(PM) and SO2 but no clear relationship with a critical gestational exposure period (Binkova et al.,
2005; Glinianaia et al., 2004a; Maisonet et al., 2004). Several subsequently published studies
revealed associations between preterm birth and 1st trimester maternal ambient air pollutant expo-
sure with weaker or no association for exposure during later pregnancy. These included studies in
Lithuania (per 10 μg/m3 increment of NO2, OR=1.69, 95% CI 1.28–2.23 [in a multipollutant
model]) (Maroziene & Grazuleviciene 2002), Australia (4th vs. 1st quartile ozone level during 1st
trimester, OR=1.26, 95% CI 1.10–1.45; similar results for NO2 and SO2) (Hansen et al., 2006) and
Korea (4th vs. 1st quartile carbon monoxide (CO) level during 1st trimester, OR= 1.26, 95% CI
1.11–1.44, p-trend < .001) (Leem et al., 2006).
A Vancouver study observed weak associations between preterm birth and ambient air
pollutant levels during the last month of pregnancy (e.g., per 1 ppm increment of CO,
OR=1.08, 95% CI 1.01–1.15) but not the first month (e.g., per 1 ppm increment of CO,
OR=0.95, 95% CI 0.89–1.01) (Liu et al., 2003). The risk of preterm birth in Taiwan was
elevated among women living close to a major freeway (<0.5 vs. 0.5–1.5 km, OR=1.30, 95% CI
1.03–1.65) (Yang et al., 2003a). In Los Angeles, there was a dose-response relationship between
preterm birth and inverse-distance-weighted traffic density among women in their 3rd trimester
during fall-winter (OR=1.15, 95% CI 1.05–1.26) but not spring-summer (Wilhelm & Ritz, 2003).
In further analysis of the Los Angeles study, preterm birth was associated with 1st trimester CO
concentrations near the maternal residence (women <1.6 km from monitoring station, per
1 ppm CO increment during 1st trimester, OR= 1.10, 95% CI 1.01–1.20; similar results for 3rd
trimester CO); preterm birth was not related to PM with a mass median aerodynamic diameter
<2.5 μm (PM2.5) concentrations during early or late pregnancy (Wilhelm & Ritz, 2005). In a
California-wide case-control study, preterm birth was associated with ambient air PM2.5 levels
near the maternal residence during early or late pregnancy (4th vs. 1st quartile PM2.5 levels
during 1st gestational month, OR=1.21, 95% CI 1.12–1.30, adjusted for CO levels; similar
results for PM2.5 during late gestation) (Huynh et al., 2006). The latter study revealed similar asso-
ciations for CO levels during early or late pregnancy, independent of PM2.5. A time-series analysis
of daily preterm birth counts in 4 Pennsylvania counties during 1997–2001 revealed associations
with PM with a mass median aerodynamic diameter <10 μm (PM10) (per 50 μg/m3 increment,
OR=1.07, 95% CI 0.98–1.18) and SO2 levels (per 15 ppb increment, OR=1.15, 95% CI 1.00–1.32)
during the 6 wk before birth (Sagiv et al., 2005). In the latter study, the associations were some-
what stronger for PM10 and SO2 concentrations 2–5 d before birth, suggesting possible acute
effects of such exposure.
Maternal exposure, industrial emissions, inadequate evidence: Retrospective cohort studies in
Taiwan revealed weak associations between preterm birth and maternal residence within 3 km of a
major oil refinery (OR=1.41, 95% CI 1.08–1.82) (Lin et al., 2001a), within 2 km of a Portland
cement plant (OR= 1.30, 95% CI 1.09–1.54) (Yang et al., 2003c), within 2 km of an industrial
complex including petrochemical, petroleum, steel, and shipbuilding industries (OR= 1.11, 95% CI
1.02–1.21) (Tsai et al., 2003) or within 3 km of coal-based electricity-generating stations
(OR=1.14, 95% CI 1.01–1.30) (Tsai et al., 2004).
Drinking-water disinfection by-products Maternal exposure, inadequate evidence: Reviewers
found inadequate evidence for an association between preterm birth and THMs (Bove et al., 2002;
Graves et al., 2001). Among reviewed studies, there were weak associations between preterm birth
and THMs in retrospective cohort studies in Sweden (hypochlorite-treated vs. unchlorinated water,
OR=1.09, 95% CI 1.01–1.17) (Kallen & Robert, 2000) and Taiwan (chlorinated vs. unchlorinated
water supply, OR=1.37, 95% CI 1.20–1.56) (Yang, 2004) but not in Denver (Gallagher et al.,
1998), Nova Scotia (Dodds et al., 1999) and Norway (Jaakkola et al., 2001b). In a recent retrospec-
tive cohort study in Massachusetts, preterm birth was inversely associated with THM levels during
the 3rd trimester in the municipality of maternal residence (≥90th vs. <50th percentile, OR=0.88,
95% CI 0.81–0.94) but not with 3-chloro–4-(dichloromethyl)–5-hydroxy–2(5H)-furanone (MX),
mutagenic activity or haloacetic acid levels (Wright et al., 2004).
Drinking water nitrate Maternal exposure, inadequate evidence: A moderately large case-control
study in Prince Edward Island, Canada, revealed an association between preterm birth and median
398 D. T. WIGLE ET AL.
well water nitrate level in the region of maternal residence at birth (≥3.1 vs. ≤1.3 mg/L, OR= 1.91,
95% CI 1.47–2.46, adjusted for prenatal smoking and several other potential confounders)
(Bukowski et al., 2001). Although suggestive, this finding requires confirmation. Groundwater nitrate
levels in rural areas may serve as proxies for pesticides and other contaminants (Ritter, 1990).
Hazardous waste sites Maternal exposure, inadequate evidence: Preterm birth was not associ-
ated with maternal residential proximity to any of 1221 NPL sites in the United States (≤1.6 vs.
>1.6 km, OR=0.99, 95% CI 0.86–1.16) (Boyle et al., 2004; Sosniak et al., 1994). In an Alaskan
study, maternal residence in villages with hazardous waste dumpsites was associated with a statisti-
cally nonsignificant increased risk of preterm birth (OR=1.24, 95% CI 0.89–1.74) (Gilbreath &
Kass, 2006b). Among studies of large single landfill sites, preterm birth risk was elevated in Los
Angeles (high-odor vs. unexposed region, mean difference in gestation length –1.8 d, p=.02)
(Kharrazi et al., 1997) and New Jersey (≤1 km and downwind vs. >1 km, high exposure period,
OR=2.10, 95% CI 1.01–4.36) (Berry & Bove, 1997) but not in Montreal (<4 km vs. unexposed
region, OR=0.97, 95% CI 0.88–1.07) (Goldberg et al., 1995). An ecologic study in Nova Scotia
revealed slightly increased risks of preterm birth in Sydney (RR=1.10, 95% CI 0.98–1.26) and the
rest of Cape Breton County (RR=1.13, 95% CI 1.04–1.22) (compared to the rest of Nova Scotia)
although the former is the site of a major hazardous waste site (Dodds & Seviour, 2001).
Solvents Maternal exposure, chlorinated solvents, inadequate evidence: A review of five epi-
demiologic studies published during 1990–2000 found inadequate evidence for an association
between preterm birth and prenatal residence in regions served by drinking water contaminated by
chlorinated solvents such as trichloroethylene (Bove et al., 2002). In a retrospective study at Camp
Lejeune, preterm birth was not associated with exposure to tetrachloroethylene-contaminated
drinking water (OR=1.1, 95% CI 0.9–1.3) (Sonnenfeld et al., 2001).
Maternal exposure, various and unspecified solvents, inadequate evidence: In a large U.S.
population-based case-control study, preterm birth was not associated with prenatal occupations
with likely exposure to benzene, petroleum, or alcohols/glycols (inferred from job titles) (Savitz
et al., 1989a). A small retrospective cohort study in Wisconsin observed no association between
preterm birth and prenatal occupations likely exposed to solvents (OR= 1.2, 95% CI 0.4–3.1)
(Hewitt & Tellier, 1998). A very small cohort study in Canada reported an elevated risk of preterm
birth among women with self-reported 1st trimester occupational organic solvent exposure (crude
OR=3.26, 95% CI 0.78–15.7, calculated from data in paper) (Khattak et al., 1999).
Paternal occupational exposure, various and unspecified solvents, inadequate evidence: A ret-
rospective cohort study based on Washington State birth certificates reported no association
between preterm birth and paternal employment in occupations with likely solvent exposure (e.g.,
painters, OR=1.0, 95% CI 0.8–1.3) (Daniell & Vaughan, 1988). A large U.S. population-based
case-control study reported no association between preterm birth and paternal occupations likely
exposed to benzene, petroleum or alcohols/glycols (inferred from job titles) (Savitz et al., 1989a).
Summary Epidemiologic evidence for the role of environmental toxicants in preterm birth
includes: (a) sufficient evidence—prenatal active smoking, ETS exposure; (b) limited evidence—
prenatal exposure to lead, PCBs (occupational exposure), DDT/DDE, outdoor air pollutants.
Fetal growth Deficit Intrauterine growth restriction (IUGR) is defined as a liveborn infant
below the 10th percentile of birth weight for gestational age. Other indicators of intrauterine growth
deficits include term low birth weight (birth weight <2500 g after at least 37 wk of gestation) and
low birth weight adjusted for gestation length. Fetal growth deficit (FGD) is defined here to encom-
pass the above indicators and is associated with increased fetal and infant morbidity and mortality.
For context, prenatal smoking is the best proven preventable cause of FGD and is responsible for
30–40% of affected infants in Canada (Health Canada, 2003).
Lead Maternal exposure, limited evidence: FGD was generally associated with maternal but
not paternal blood lead levels (Figure 2). Two birth cohort studies conducted in lead smelter towns
revealed no association between FGD and prenatal blood, cord blood, or placental lead levels
(Loiacono et al., 1992; McMichael et al., 1986). A birth cohort study in Boston reported a dose-
response relationship between FGD and cord blood lead levels (relative risk increment per unit
increase in cord blood lead concentrations [μg/dl], 1.06, 95% CI 1.00–1.13) (Bellinger et al., 1991).
CHILD HEALTH AND ENVIRONMENTAL CONTAMINANTS 399
Several recent cohort studies have consistently reported significant associations between FGD and
maternal lead exposure, including: (1) tibial bone lead (change in birth weight per unit increase in
bone lead level [μg/g] adjusted for gestation length, –7.29 ± 2.45 g, p=.003) (Gonzalez-Cossio
et al., 1997), (2) maternal occupations likely exposed to lead (OR = 1.34, 95% CI 1.12–1.60) (Irgens
et al., 1998), OR=2.8, 95% CI 0.8–9.6 (Seidler et al., 1999), (3) cord blood lead (change in birth
weight per unit change in cord blood lead, –9.7 g, 95% CI –16.9 to –2.5) (Osman et al., 2000),
(4) placental lead concentration (change in birth weight per 0.1 μg/g placenta lead increment adjusted
for gestation length, –73.6 g, 95% CI –152.7 to 5.5) (note that the range of placenta lead concentrations
was 0.03–0.57 μg/g) (Odland et al., 2004) and (5) maternal blood lead level (10–19 μg/dl, OR=1.62,
95% CI 0.91–2.75; ≥20 μg/dl, OR=2.15, 95% CI 1.15–3.83, p-trend < .01 (Chen et al., 2006); maxi-
mum prenatal blood lead ≥10 vs. <10 μg/dl, OR=4.2, 95% CI 1.3–13.9 (Jelliffe-Pawlowski et al.,
2006). Maternal exposures to airborne lead emissions in Shoshone County, Idaho (during a 15-mo
period when air emissions were high because of a damaged bag house) was associated with FGD
(OR=1.92, 90% CI 1.33–2.76) (Berkowitz et al., 2006).
Paternal occupational exposure, inadequate evidence: A case-control and two retrospective
cohort studies revealed no association between FGD and paternal employment in jobs likely
exposed to lead (Irgens et al., 1998; Kristensen et al., 1993; Savitz et al., 1989a). Two retrospective
cohort studies in New York State and Taiwan showed no association between FGD and a history of
blood lead levels above 25 μg/dl for at least 5 yr before conception (OR=0.82, 95% CI 0.28–2.37)
(Lin et al., 1998) or preconceptual blood lead level of at least 20 μg/dl (OR=0.94, 95% CI 0.51–1.62)
(Chen et al., 2006).
Inorganic arsenic Maternal exposure, airborne, inadequate evidence: A U.S. case-control
study of FGD infants found no association with self-reported prenatal occupational exposure to
airborne arsenic (OR=0.8, 95% CI 0.4–1.5) (Savitz et al., 1989a).
Maternal exposure, drinking water, inadequate evidence : A birth cohort study in Chile
revealed reduced birth weight among infants of women living in a city with average drinking water
arsenic levels of 40 μg/L; birth weight adjusted for gestation length was 57 g less than that in a city
FIGURE 2. Fetal growth deficit vs. parental lead exposure (m =prenatal, p=paternal, Pb=lead, BPb= blood lead, occup=occupation).
Huel
Bellinger
Irgens
Seidler
Berkowitz
Jelliffe
Chen
Savitz
Kristensen
Irgens
Lin
Chen
Study
m
m
m
m
m
m
m
m
p
p
p
p
p
p
Grp
hair Pb > 14
BPb inc = 10
occup
occup
airborne Pb
BPb 10+
BPb 10–19
BPb 20+
occup
occup
occup
BPb 25+
BPb 10–19
BPb 20+
Compare
1981
1991
1998
1999
2006
2006
2006
1989
1993
1998
1998
2006
Yr
0.1 0.2 0.5 12 5
10
Odds ratio and 95% CI
400 D. T. WIGLE ET AL.
with drinking-water arsenic levels below 1 μg/L (95% CI–123, 9) (Hopenhayn et al., 2003). How-
ever, the association between birth weight and individual tap water arsenic levels (β=–0.26 g, 95%
CI –0.85 to 0.31, per μg/L) was statistically nonsignificant.
Paternal occupational exposure, airborne, inadequate evidence: A U.S. case-control study of
FGD infants found no association with self-reported paternal occupational arsenic exposure
(OR=1.2, 95% CI 0.8–1.8) (Savitz et al., 1989a).
Cadmium Maternal exposure, inadequate evidence: In a small retrospective cohort study,
there was a statistically nonsignificant inverse association between FGD and maternal hair cadmium
(≥0.42 vs. <0.42 μg/g, OR= 1.68, 95% CI 0.35–8.34); this study did not adjust for prenatal smok-
ing (an important source of cadmium exposure and a known cause of FGD) (Huel et al., 1981).
Two birth cohort studies of nonsmoking pregnant women found no association between birth
weight adjusted for gestation length and maternal or cord blood or placental cadmium levels (Kuhnert
et al., 1987; Zhang et al., 2004). A very small Italian birth cohort study with limited statistical analy-
sis reported an inverse correlation between birth weight of term infants and maternal blood cad-
mium (Pearson’s r=–0.55, p=.0003) (Salpietro et al., 2002). A small birth cohort study in a
cadmium-polluted region of Japan observed no association between birth weight and maternal
urinary cadmium levels (Nishijo et al., 2002), but there was an inverse relationship between height
at birth and 3rd trimester maternal blood cadmium (β=–0.59 ± 0.277 cm, p=.04) (Nishijo et al.,
2004). A recent Norway/Russia birth cohort study reported no association between birth weight
and maternal blood cadmium level (Pearson’s r=–0.23, p > .05); this study did not adjust for ges-
tation length but only 2 of the 55 infants were preterm (Odland et al., 2004). Reviewers concluded
that high-dose prenatal cadmium exposure causes fetal growth deficits in experimental animals, but
there was little evidence for a relationship at the much lower exposure levels observed in humans
(Agency for Toxic Substances and Disease Registry, 1999a).
PCBs Prenatal occupational exposure, inadequate evidence: Among infants of women occu-
pationally exposed to airborne PCBs, there was an inverse association of borderline statistical signif-
icance between birth weight adjusted for gestation length and prenatal serum PCB levels estimated from
those measured in a sub-sample of women (per 2.7-fold maternal serum PCB increment, β=–24 g,
90% CI –49 to 2) (Taylor et al., 1989). A German birth cohort study found no association between
FGD and prenatal PCB exposure inferred from a job-exposure matrix (exposed vs. unexposed,
OR=1.2, 95% CI 0.8–1.7) (Seidler et al., 1999).
Prenatal environmental exposure, inadequate evidence: Reviewers found inadequate evi-
dence for an inverse association between birth weight and maternal exposure to background
environmental PCBs (Longnecker et al., 1997). Studies published since this review provide incon-
sistent evidence for an association. In a study of Swedish fishermen wives, low birth weight was
associated with prenatal serum PCB-153 levels (>400 vs. ≤400 ng/g lipid, OR=2.3, 95% CI 0.9–
5.9), but there was no adjustment for gestation length (Rylander et al., 1998). Birth weight
adjusted for gestation length was inversely associated with cord plasma PCB levels in Holland (per
2.7-fold plasma PCB increment, β=–119.4 ± 53.7 g, p=.03) (Patandin et al., 1998). There was
an increased risk of FGD (borderline statistical significance) among Swedish fishing families in a
region where fish had relatively high PCB levels (contaminated vs. less contaminated region,
OR=1.4, 95% CI 0.9–2.1) (Rylander et al., 2000). A retrospective cohort study in New York State
found a weak but statistically significant association between low birth weight (adjusted for gestation
length and other potential confounders) and prenatal residence in regions with PCB-contaminated
hazardous waste disposal sites (OR=1.04, 95% CI 1.02–1.07) (Baibergenova et al., 2003). In a
retrospective cohort study of Lake Michigan female anglers, birth weight adjusted for gestation
length was reduced among women in the highest serum PCB category (serum PCB 25–29 vs.
<5 μg/L, mean birth weight 2958 ± 224.0 vs. 3520 ± 103.3 g,