Off to a good start: the influence of pre- and periconceptional exposures, parental fertility, and nutrition on children's health.

The scientific community is developing a compelling body of evidence that shows the importance of the in utero environment (including chemical and hormonal levels) to the ultimate health of the child and even of the aging adult. This article summarizes the evidence that shows this impact begins with conception. Only a full life-cycle evaluation will help us understand these impacts, and only such an understanding will produce logically prioritized mitigation strategies to address the greatest threats first. Clearly, the time for analysis begins when the next generation is but a twinkle in the eye.

There is a growing appreciation of the impact that very early life environment (birth weight, trauma, xenobiotic exposures) has on the health of the child and the adult. The lifecourse approach to health and disease recognizes that these very early impacts contribute to disease later in life (reviewed in Gillman 2002). From this perspective, any long-term prospective study that attempts to uncover and identify effects on children's health should include an evaluation of the health of both parents prior to and at the time of conception, including xenobiotic load, nutritional state of the mother, and the involvement of any assisted reproductive technologies. These features have already been shown to have demonstrable short-term impacts on the health of the child. Long-term follow-up is essential if we are to fully understand the health burden attributable to these features and to develop more informed hypotheses about exposure-disease relationships. Only from such a complete picture can we hope to make policy decisions that will use our resources most effectively to improve the health of generations to come.
Although by no means an exhaustive review, this article presents some of what we know about these impacts. We conclude that explicit assessment of such exposures and factors before and around the time of conception is indispensable for a real understanding of the determinants of health in our children, and, by extension, the next generation of adults.

The Case for Early Analyses
Explicit measurements of exposure, body burden, or nutritional status in a couple before conception is necessary for numerous reasons.
One critical reason is that recall is variably dependable. Some studies found that the ability of a person to recall specific exposures was good to excellent for some medications (reviewed in Harlow and Linet 1989;Kelly et al. 1990) and some environmental exposures and pesticides (Blair et al. 2001;Feldman et al. 1989) but no better than chance for others (Feldman et al. 1989;Kelly et al. 1990). Although multiple questionnaires may increase the likelihood of obtaining an accurate picture of what the exposures truly were (Farrow et al. 1996), this can be cumbersome and may yield a relatively small increase in confidence.
Additionally, questionnaires have a significant limitation: they capture only what is remembered, which can come only from what the participant knows or is aware of. Importantly, environmental exposures are occult. Exposures to and absorption of excipients and active agents from manufactured products frequently occur without our knowledge or awareness. Phthalates in air or cosmetics can lead to measurable levels in vivo, for example (Blount et al. 2000), an internal exposure that no questionnaire, however well constructed and administered, would be able to find and reconstruct. Specific phthalates are capable of disrupting reproductive development in rodents (reviewed in CERHR 2002) and are suspected of causing similar effects in other species, so this gap in our understanding is critical. The limitations of questionnaire data were demonstrated recently for polychlorinated biphenyls (PCBs), where there was no association between potential sources of exposure by questionnaire and actual serum levels of several coplanar PCBs (Shadel et al. 2001). Because few consumers are aware of their exposures or the sources thereof, questionnaire data will not be useful for categorizing study subjects into most-exposed and least-exposed groups for many compounds of current (and probably future) concern.
Arguably, one of the best reasons for performing a prospective longitudinal study is the high probability of finding new associations between previously unanticipated exposures and a health outcome. The more we know about what compounds are in the body and in what amounts, the greater our confidence will be in associating those exposures with health outcomes. This accuracy cannot be acquired with questionnaire proxy data, but requires actual measurements as well as measurements of compounds beyond those that we know today to be hazards.
Making concurrent measurements of body burden is important because of the changing nature of xenobiotics. One of the results of phasing out persistent organic pollutants is that more xenobiotics now have shorter half-lives, and a biological sample taken in late pregnancy or after birth cannot be considered representative of in vivo levels at the time of conception or during early pregnancy (reviewed in MacIntosh et al. 1999). An excellent example is methoxychlor, developed to replace dichlorodiphenyltrichloroethane (DDT). It has a similar spectrum of intended effects but is much more readily excreted and thus has a much-reduced tendency to bioaccumulate (Kapoor et al. 1970). Phthalates are another example: these are cleared from the body with relative speed (several days to a few weeks) and do not bioaccumulate significantly (Tanaka et al. 1978). From the public health and environmental perspectives, this is incontrovertibly good, although it means that actually determining which levels of which compounds are adverse requires repeated biological sampling and analysis.
A corollary of these short half-lives is considerable sample-to-sample variation. Even with no obvious changes in behavior, urinary levels of some pesticides have been shown to vary by factors of four times or more (MacIntosh et al. 1999). Consequently, a single sample will be poorly representative of the long-term levels in that individual.
A benefit of repeated sampling is that it would accommodate the altered behaviors exhibited by women when they know they are pregnant (Lelong et al. 2001;Root and Browner 2001). If these altered behaviors also reduced their exposures to, and internal burdens of, harmful xenobiotics, then the reduced internal half-lives plus the altered behaviors will significantly reduce internal levels, making late-stage analyses even less representative of earlier levels. Measuring internal levels of xenobiotics around the time of conception will provide a much more realistic picture of the exposure of the embryo at the critical period of conception.
Study of the nutritional status of a woman prior to pregnancy would allow for the timely measurement of prepregnancy weight and the collection of dietary history and laboratory data. Studies have found that self-reported prepregnancy weight may be biased, especially among overweight women, who are more likely to underestimate their prepregnancy weight (Perry et al. 1995;Stevens-Simon et al. 1986). Dietary deficiencies, thought to be due to long-standing dietary habits and not a sudden insult, may resolve if a woman changes her diet after she knows that she is pregnant. By getting a dietary history prior to conception, we can eliminate recall bias and document diet during the critical periods (Sempos 1992;Wei et al. 1999). Early collection of laboratory data is important for several reasons. First, hemodilution, which begins early in the second trimester and varies in degree and timing among individuals, makes interpretation of a number of micronutrient levels difficult (e.g., plasma zinc concentrations). Second, other physiologic changes that occur during pregnancy (e.g., increased renal excretion of water-soluble vitamins, hormone-mediated transfer of nutrients to placenta and mammary glands, decrease of serum albumin) make determining micronutrient status during pregnancy difficult (King 2000). Third, it may be important to measure micronutrient levels prior to pregnancy because changes in behavior (e.g., cessation of smoking and/or alcohol consumption) and use of iron and multivitamin supplements during pregnancy may affect laboratory values. Last, changes in the type and quantity of food consumed as well as changes in physical activity may affect values (King 2000). By evaluating nutritional status prior to pregnancy or in early pregnancy prior to physiologic and behavioral changes, one can more accurately evaluate a woman's nutritional status and the associations between nutritional status during early pregnancy and subsequent functional outcomes. Knowledge of such associations could lead to interventions that prevent adverse outcomes, such as the provision of folic acid to prevent birth defects.
The rest of this review will focus on a variety of key issues and their relationships to the health of the resultant child: compounds in semen; compounds in the prepregnant female; the effects of assisted reproductive technologies (ARTs); the effects of preimplantation exposures; the influence of early maternal nutrition; and a review of the period of greatest sensitivity for producing malformations and fetal alterations.

Compounds in Semen
The pathway through which environmental toxicants contaminate semen and affect offspring is diagrammed in Figure 1. In this scenario, a male is exposed to an environmental toxicant that is absorbed, distributed, biotransformed, and excreted as metabolite or parent compound in seminal fluid. Alternatively, compounds can be adsorbed to sperm and be introduced directly into the egg at fertilization [e.g., cocaine or cyclophosphamide (reviewed in Hales and Robaire 2001;Yazigi et al. 1991)]. Any contaminants in semen are transmitted to a woman in the ejaculate. The contaminant is absorbed by the female, where it may reach and adversely affect a current pregnancy and perhaps remain in the woman's body to influence future pregnancies, or it enters the egg with the sperm and provides an initial direct dosing of the conceptus.
Metals such as lead and cadmium, both known developmental toxicants, have been measured in human seminal fluid by a number of research groups (Dawson et al. 2000;Kumar et al. 2000;Noack-Fuller et al. 1993;Saaranen et al. 1989;Stachel et al. 1989;Telisman et al. 2000;Xu et al. 1993). The organic solvents benzene, toluene, and xylene can be found in semen (Xiao et al. 1999), and ethanol can be detected, entering through simple diffusion (Luke and Coffey 1994). Finally, seminal fluid of smokers contains nicotine, its metabolites, aromatic hydrocarbons, and precursors of mutagenic nitrosamines (Hoffman et al. 1994;Pacifici et al. 1993;Rivrud 1988;Wong et al. 2000). The critical question of dose sufficiency (are these compounds present at a level sufficient to cause adverse effects?) remains unanswered.
The composition of seminal fluid changes normally as a function of frequency of ejaculation, hydration, nutritional state, and exposures to the male. Because of this variability, it would be necessary to monitor the fluid in a prospective way to make clear associations between contaminants present in seminal fluid (and transmitted to a female partner) and any abnormalities in the offspring. Only recently have laboratories been capable of measuring multiple low-concentration contaminants in ever-smaller amounts of seminal fluid. This capability presents a significant opportunity. A large longitudinal study, coupled with advances in detection and measurement of chronic, low-level contaminants in seminal fluid, would provide an excellent opportunity to more clearly delineate any associations between semen-mediated very early exposures and health outcomes in the offspring.

Altered Semen Quality
An important area often neglected when evaluating environmental exposures and children's health relates to sperm DNA-derived effects. Exposures to the father that cause germline DNA or chromatin damage and consequently affect offspring health are well documented in animal models (reviewed in Hales and Robaire 2001). Similar studies in humans (exposing the male to toxicants, mating over time, monitoring offspring for adverse effects) cannot even be approached for obvious ethical reasons. Yet, there is undisputed evidence that the human male significantly affects the health of his children through sperm DNA in paternally mediated Mendelian genetic defects, chromosomal aberrations (e.g., paternal translocation carriers), and aneuploid sperm (associated with 30-50% of Klinefelter, 80% of Turner, 5% of Down, and 100% of extra Y chromosome syndromes) (Blanco et al. 1998;Eskenazi et al. 2002;Hassold 1998;Hixon et al. 1998;Martinez-Pasarell et al. 1999;Pangalos et al. 1992). Researchers in ART clinical settings have shown that both paternal sperm chromatin and DNA quality affect pre-and postimplantation development of offspring (Evenson et al. 2002;Hammadeh et al. 1996;Tesarik et al. 2002). It is evident that sperm with abnormalities of DNA and/or chromatin are able to fertilize and transmit abnormal DNA to the conceptus. It is also evident from numerous studies that environmental exposures can induce mutations, chromosomal aberrations, chromatin, or epigenetic protein changes in human sperm (selected examples in Table 1). This leads to a model for environmentally induced male-mediated effects on offspring, as depicted in Figure 2. This model clearly demonstrates the importance of a study design that monitors environmental exposures to the father during sperm development and maturation prior to and at the time of fertilization. Because sperm are constantly being replenished in the male genital tract, only prospective study designs are able to capture direct effects of environmental exposures on sperm DNA quality at the time of conception. Previous studies that have tried to establish cause-effect relationships by evaluating DNA quality of sperm after the birth of an abnormal child have not been successful because of this temporal flaw. Recent advances in molecular-cell and genetic laboratory techniques now allow sophisticated assessment of human sperm DNA and chromatin damage (Evenson et al. 2002;Perreault et al. 2000Perreault et al. , 2003. There are new understandings of the critical role of sperm chromatin organization and paternal DNA in early embryo development (Evenson 1999;Sakkas et al. 1998;Tesarik et al. 2002). A prospective longitudinal study would be the most definitive way to finally answer the question of dose, timing, and types of exposures that cause specific sperm DNA and chromatin damage that affects offspring health.

Compounds in the Female
In addition to the father's contribution, the mother brings a time-averaged and complex body burden of xenobiotics to the conception and the early pregnant period.
What is known about the effects of compounds at this time is recent and relatively limited. This is partly because earlier studies examining the impact of these exposures in the environmental setting and reproductive effects were severely hampered by crude measures of exposure. Lack of biological assays to measure many of these contaminants in tissue specimens, the high cost of using these assays in large epidemiologic studies when they were available, and challenges related to the interpretation of the results when exposures were measured after the adverse reproductive event occurred limit the conclusions that can be drawn from these investigations. Rapid advancements in laboratory techniques, including biological assays as screening methods for exposure to these contaminants, as well as highly sensitive techniques to measure specific congeners of the compounds or their metabolites at reasonable costs are exciting developments that hold great promise in enhancing our understanding of the impact of these exposures on reproductive events Seidel et al., 2000).
Despite these impediments, data do exist from national studies to describe general population exposures to some of these contaminants. The National Health and Nutrition Examination Survey II (NHANES II) measured serum levels of selected pesticides in approximately 21,000 participants ranging in age from 12 to 74 years (Murphy and Harvey 1985). Increasing age was related to the presence of detectable levels of DDT, with the proportion exposed ranging from 14 to 51% across the age spectrum. Serum values ranged from 2 to 58 ppb. Detectable levels of the DDT metabolite p,p´-DDE were found in serum in nearly 99% of the population, with values ranging from 1 to 378 ppb, which also increased with age.
Although the use of DDT in the United States has been banned for more than 30 years, exposure may continue through consumption of contaminated foods, including products imported from countries where DDT is still being used, as well as foods contaminated because of the environmental persistence of DDT. Additional data on serum DDE levels in selected subgroups from the U.S. population are available from several investigations that were conducted to examine the effect of DDE and PCB exposure and risk of breast cancer. A recent study involved the pooling of five U.S. studies that included women over an interval spanning 1974-1997 (Laden et al. 2001). Median values for serum DDE levels ranged from 0.41 to 1.67 µg/g lipids. Several studies have also assessed DDE levels among consumers of sport-caught fish  from contaminated waters (Hanrahan et al. 1999;Schantz et al. 2001). These reports indicate that consumers of sport-caught fish have higher serum DDE levels compared with those of nonconsumers (Hanrahan et al. 1999;Schantz et al. 2001). These same studies in consumers of sportcaught fish versus nonconsumers have provided data on serum PCB levels among consumers compared with non-fish-eaters. Background PCB levels in the U.S. population were reported to be 4-8 ppb in the early 1980s (Kreiss 1985) and is currently 0.9-2.2 ppb (ATSDR 2000). In the Hanrahan study conducted in 1993-1994, male fish-eaters had a mean serum PCB level of 4.8 ppb compared with 1.5 ppb among their nonconsuming counterparts (Hanrahan et al. 1999), whereas female fish-eaters had levels of 2.1 ppb versus 0.9 ppb among female non-fish-eaters. Participants in the Michigan fish-eaters cohort, which consists of individuals who were consuming ≥ 26 pounds of sport-caught fish per year (fish-eaters) and those eating < 6 pounds per year (non-fish-eaters), when enrolled in 1980, had considerably higher serum PCB levels (Tee et al. 2003).
How do these maternal burdens impact the developing fetus or child? In a cohort of infants born to Michigan women in 1983 that included consumers of sport-caught fish and those who ate fish other than sport caught, significantly lower birth weight and delayed motor reflexes were noted among exposed infants at birth (Fein et al. 1984). Overall, mean PCB levels in cord blood PCB in this study were 3 ng/mL (± 2 ng/mL), and in maternal serum 6 ng/mL (± 4 ng/mL). Importantly, persistent effects on cognitive functioning among the exposed children were reported in a follow-up at 11 years of age (Jacobson and Jacobson 1996). This increased burden tends to weigh most heavily on lower socioeconomic minority populations (Sweeney et al. Unpublished data), and these increased exposures bring adverse consequences. Only by determining with certainty the different exposures during development and following many outcomes postnatally can we ascertain the full burden of these exposures, which is the first step toward their elimination.
Other examples of endocrine-active compounds (EACs) that have been detected in serum and adipose tissue samples obtained from representative populations from the United States include beta-benzene hexachloride, dieldrin, trans-nonachlor, hexachlorobenzene, heptachlor epoxide, and oxychlordane (Murphy and Harvey 1985). Another class of EACs measured in NHANES III, the phthalates, has generated considerable concern regarding potential adverse reproductive effects. Using a novel and highly selective technique, investigators measured monoester metabolites of seven commonly used phthalates in 289 adults from the NHANES III study population (Blount et al. 2000). An important finding of this study was the significantly higher urinary levels of monobutyl phthalate among women of reproductive age compared with those of other age/gender groups. In addition, rural women of childbearing age had higher levels of benzyl butyl phthalate compared with those of older rural women. This is of concern because both of these phthalate isomers have been implicated in animal studies of reproductive toxicity (Gray 1998;Jobling et al. 1995;Shiota et al. 1980;Wilkinson and Lamb 1999). Human studies are scarce, but a recently published case-control study examined the association between premature thelarche (premature breast development) and phthalate exposure among girls 6 months to 8 years of age (Colon et al. 2000). With serum as the medium, significantly higher levels of dimethyl, diethyl, dibutyl, and di-2-ethylhexyl phthalate (DEHP) were measured in the cases. With regard to DEHP, the ratio of average concentrations between control:case samples was 70:450 ppb. Animal studies show that DEHP has its greatest effects while the reproductive system is forming (within the first 4 weeks in a human pregnancy), so identifying and quantifying phthalate levels prepregnancy will be particularly important in ensuring that all the impacts on reproductive development are accounted for and quantified.
Finally, although not classified as EACs, lead and mercury are important environmental contaminants associated with adverse effects on reproduction. Although an elevated lead level in children is defined as > 10 µg/dL by the Centers for Disease Control and Prevention (CDC 2000), a "safe" lead level remains controversial (Sanborn et al. 2002). As with PCB levels, blood lead levels in the United States have been steadily declining. However, subgroups, particularly low-income populations and children requiring social services and foster care, remain at high risk for elevated blood lead levels (Chung et al. 2001).
Organic mercury, which is mostly methyl mercury, bioaccumulates in the aquatic food chain and can be present in relatively large amounts in fish living in contaminated waters. Effects on the developing nervous system have been well documented in studies following the Minamata Bay and Iraq poisoning incidents (reviewed in CDC 2000), as well as in more recent studies of children exposed prenatally at much lower levels (Grandjean et al. 1998). A recent report issued by the National Research Council estimates that 60,000 infants per year in the United States may be prenatally exposed to levels of methyl mercury sufficient to cause neurological and cognitive impairments (National Research Council 2000).
Concern has also focused (Toppari et al. 1996) on the possible role of endocrinedisrupting chemicals with either estrogenic or antiandrogenic effects in the marked rise (almost doubling) in the prevalence of hypospadias in the United States and Europe since the late 1960s (Kallen et al. 1986;Paulozzi et al. 1997). Because the development of the male genital organs is under hormonal control, the main hypotheses as to the etiology of hypospadias are centered on either disturbance of endogenous hormonal production or exposure to exogenous hormones (Dolk 1998). The hypothesis is supported by the most recent findings of an increased risk of hypospadias in sons of women who themselves had been prenatally exposed to diethylstilbestrol (DES; a synthetic estrogen widely prescribed to pregnant women between 1938 and 1975 to treat habitual abortions) (Klip et al. 2002). This transgenerational effect was associated with maternal in utero exposures to DES early in the first trimester. A link with cryptorchidism has also been noted. Because the reproductive system forms early in the first trimester, obtaining the best possible estimates of exposure during that time will help us determine the degree of environmental contribution to these problems.
Early first-trimester drug exposures have been linked to birth defects. Recent examples are the illicit use of amphetamines associated with significantly increased risk of predominantly cardiovascular and musculoskeletal malformations (McElhatton et al. 1999), and the misuse of misoprostol, a synthetic analog of prostaglandin E, resulting in limb deformities attributable to vascular disruption (Gonzales et al. 1998).
Adverse developmental outcomes have also been associated with early exposures to environmental physical factors. Pre-and periconceptional parental occupational exposures to low-level ionizing radiation have been repeatedly associated with increased risk of fetal loss and neural tube defects (NTDs) (Doyle et al. 2000;Inskip 1999;Sever et al. 1988).
Given the findings above and the recent quantitation of a number of xenobiotics in human ovarian follicular fluid (Younglai et al. 2002), the impact of these environmental chemicals on human reproduction has become a high priority for public health researchers (Barlow et al. 1999). A paper describing the conclusions reached by the recent U.S. Environmental Protection Agency-sponsored "Workshop to Identify Critical Windows of Exposure for Children's Health" called for research into the effects of environmental exposures during the periconceptional and prenatal periods on reproductive and children's health (Pryor et al. 2000). With advances in laboratory techniques and statistical modeling approaches for exposure assessment, we are now positioned to be able to quantify the degree of exposure that occurs early, which will affect strategies for preventive health policies in the future.

Assisted Reproductive Technologies
The evolution of and increasing use of various ARTs for the treatment of infertility has provided additional insight into the importance of considering exposures that occur very early in conception or even just prior to conception. ART encompasses those procedures in which both eggs and sperm are handled in the laboratory. To date, a range of effects have been reported in association with various facets of ART procedures. Some effects can be linked to specific exposures that occur in conjunction with a particular step in the ART cycle; for other effects, an association with ART is suggested, but it is not possible from available data to disentangle the specific exposures or mechanisms. Nonetheless, the data accumulating from ART pregnancies underscore the importance of periconceptional exposures.
The vast majority of ART procedures include ovarian stimulation with either clomiphene citrate, human pituitary gonadotropins (various formulations), or both. Several potential adverse effects have been suggested from studies focusing on the use of these ovulation-stimulating drugs. Perhaps the most striking finding is the association reported between artificial induction of ovulation and monozygotic (MZ) twinning. Unlike dizygotic twinning, the prevalence of MZ twinning has been remarkably constant worldwide (Dunn and Macfarlane 1996;Guttmacher 1953;Hur et al. 1995;MacGillivray 1986;Parazzini et al. 1991). However, in the late 1980s, Derom and colleagues reported a 2-fold increase in the MZ twin birth rate among women from their population-based study in East Flanders, Belgium, who had conceived after therapy with an ovulation-inducing medication, compared with the rate among women who conceived naturally (Derom et al. 1987). This was the first report that an extrinsic factor might affect zygotic division. Because twinning carries a suite of increased health risks (Kovacs et al. 1989;Naeye et al. 1978), this is not viewed as a benign outcome. Use of ovulation drugs, particularly clomiphene, has also been implicated as a risk factor for spontaneous abortion (Balen et al. 1993;Dickey et al. 1996;McFaul et al. 1993;Oktay et al. 2000;Shoham et al. 1991), preterm delivery (Olivennes et al. 1993;Sundstrom et al. 1997) and low birth weight (Olivennes et al. 1993;Sundstrom et al. 1997). This research raises the question about whether such medications might adversely affect oocytes directly. Studies in mice suggest that clomiphene might affect oocyte maturation and rates of aneuploidy (London et al. 2000).
Perhaps the most obvious source of unique exposures linked to ART pregnancies is the culture media in which preimplantation embryos are fertilized and cultured. A body of both animal and human research is accumulating that implicates several factors in culture media as having an effect on embryo development. High concentrations of glucose in culture media have been associated with decreased embryo quality, defined by cleavage rates, degree of fragmentation, and blastocyst development (Coates et al. 1999;Gardner et al. 2000). These effects were reported from studies of human embryos and embryos from several other mammalian species and may be mediated by altered intracellular pH (Bavistor 1999). Mouse embryos incubated in serumcontaining media had altered cleavage rates compared with the rates in embryos cultured in media without serum (Khosla et al. 2001). Moreover, the addition of serum to culture media has been linked to effects on DNA methylation and deregulation of imprinted genes. These findings have implications for outcomes associated with imprinted genes, such as those controlling fetal growth. Finally, several case reports and studies have suggested a general association between in vitro culture to the blastocyst stage and MZ twinning (Behr et al. 2000;da Costa et al. 2001;Gorrill et al. 2001;Peramo et al. 1999;Sheiner et al. 2001;Van Langendonckt et al. 2000).
ART embryos might also be affected by various techniques in which the embryos are manipulated beyond in vitro culture. In intracytoplasmic sperm injection (ICSI), a single spermatozoon is directly injected into the egg through the oocyte membrane. Studies on embryos from various mammalian species suggest that ICSI is related to a host of chromosomal abnormalities including effects on spindle apparatus, microtubules, chromatin behavior, cell cycle checkpoints, and chromosome positioning (Hewitson et al. 1999;Luetjens et al. 1999;Ramalho-Santos et al. 2000;Sutovsky et al. 1996;Terada et al. 2000). Increases in both de novo sex chromosome abnormalities and de novo autosomal rearrangements have been reported among infants conceived using ICSI (Bonduelle et al. 2002). The other, longer-term health changes associated with this procedure have yet to be evaluated.
Another embryo manipulation technique that has raised concern is "assisted hatching" (Alikani et al. 1994;Nijs et al. 1993;Skupski et al. 1995;Slotnick and Ortega 1996;). Comparative studies have documented an increased risk for MZ twinning in ART pregnancies that included assisted hatching (Hershlag et al. 1999;Schieve et al. 2000), and animal studies have shown that MZ twinning can be induced through mechanical or chemical manipulation of the zona pellucida (Allen and Pashen 1984;Picard et al. 1985;Sotomaru et al. 1998;Talansky and Gordon 1988).
Infant outcomes have also been linked to ART generally. ART is recognized as an important contributor to the U.S. low birth weight rate because of the known associations between ART and multiple birth (CDC 2002a(CDC , 2002b and between multiple birth and low birth weight (Martin and Park 1999). Additionally, studies have suggested that low birth weight rates are increased among singleton infants conceived with ART compared with naturally conceived infants or population-based rates (Bergh et al. 1999;Dhont et al. 1999;FIVNAT 1995;Friedler et al. 1992;Gissler et al. 1995;MRC Working Party 1990;Schieve et al. 2002;Verlaenen et al. 1995;Westergaard et al. 1999). Additional analysis suggests that the increased risk for term low birth weight may be related to the ART procedure itself rather than the underlying infertility.
Equivocal results have been reported for the association between ART and birth defects in the offspring (Bergh et al. 1999;Bonduelle et al. 2002;Dhont et al. 1999;Ericson and Kallen 2001;FIVNAT 1995;Friedler et al. 1992;Hansen et al. 2002;MRC Working Party 1990;Verlaenen et al. 1995;Wennerholm et al. 2000;Westergaard et al. 1999). Studies to date have suffered from various methodological problems, including low statistical power, particularly to assess individual defects separately, and differential case ascertainment and coding schemes for infants conceived using ART and infants conceived naturally. Nearly all studies relied on retrospective registry data. Two recent studies have shown increased risk for various birth defects among ART-conceived infants (NTDs, alimentary tract atresia, omphalocele, hypospadias, and cardiovascular, urogenital, chromosomal, and musculoskeletal defects) (Ericson and Kallen 2001;Hansen et al. 2002). These studies are particularly noteworthy because they demonstrated elevated risks even among singleton infants. A longitudinal prospective study that looks at long-term outcomes would be immensely valuable in delineating yet-undiscovered associations between long-term health and method of conception. These concerns are made more acute by a recent study that reported an increased risk for developmental delay and cerebral palsy among children conceived with ART (Stromberg et al. 2002). These effects remained elevated when analyses were limited to singleton births, although the study suffered from a number of methodological drawbacks, including a lack of statistical power to adequately assess subgroup findings.
Collectively, these data show that ART methods have health consequences for the offspring. Most of these studies have been

Mini-Monograph | Periconceptional exposures and children's health
Environmental Health Perspectives • VOLUME 112 | NUMBER 1 | January 2004 relatively short-term and have limited power to evaluate correlations between early procedures and exposures, and later health outcomes. Given that this much is already known about health effects of ART, we expect that more relationships would be found if we evaluated health over a longer term. A large longitudinal study would be uniquely suited to define these relationships and would help prospective parents to make fully informed decisions about the true costs and burdens of ART methods.

Preimplantation and Early Gestational Exposures
Implantation in humans is a gradual process that takes up most of the second week of pregnancy (reviewed in O'Rahilly and Muller 2001). In rats and mice, implantation occurs between days 4 and 6 (reviewed in Cummings 1993). Although organogenesis was believed to be the most vulnerable time for inducing terata (see below), some recent studies have found that preimplantation exposures can have serious impacts on fetal health.
The first example of this, ethylene oxide, was reported by Generoso et al. (1987). Subsequently, that same group reported that a number of chemicals are potent teratogens when administered to mice 1-6 hr after mating but are ineffective either before or after that window (reviewed in Rutledge et al. 1992). In addition, the microtubule disruptor nocodazole reportedly can disrupt midgestation development when administered only at the time of sperm entry (Generoso et al. 1989). The authors conclude that there are primarily epigenetic mechanisms at work.
Subsequently, cyclophosphamide, actinomycin D, methyl mercury, alkylating agents, cross-linkers, and direct and indirect mutagens have all been found to cause malformations or growth alterations in offspring when administered to pregnant rodents prior to implantation (Iannaccone et al. 1987;reviewed in Rutledge et al. 1992).
Less extreme exposures have effects in humans. Parental periconceptional smoking, for example, was associated with a significant decrease in offspring male:female ratio. In a study on 11,815 singleton infants delivered in Japan (Fukuda et al. 2002), the daily parental consumption of cigarettes during the periconceptional period (from 3 months before the last menstruation to when the pregnancy was confirmed) was associated with a significant decline in the neonatal male:female ratio.
This study fits into the larger picture of the male:female newborn ratio decline in a number of industrialized countries during the past decades, including Denmark, the Netherlands, Norway, Finland, Sweden, Germany, Canada, and the United States (Davis et al. 1998). The reason for this reduction is not clear, but it has been suggested that chronic exposures to toxic environmental agents that predominantly affect males and the male reproductive system could lead to a lower male:female ratio, as seen, for example, after the exposures to dioxins in Seveso, Italy (Mocarelli et al. 2000), PCBs and dibenzofurans in Taiwan (Rogan et al. 1999), DBCP in Israel (Potashnik et al. 1984), medical radiation (Hama et al. 2001), and methyl mercury in Minamata Bay (Sakamoto et al. 2001). Evidence has emerged that mammalian sex ratio at birth is partially controlled by parental hormone levels at the time of conception (James 1996). These levels could be altered by exposures to endocrine-active substances such as exogenous estrogens or by compounds that interfere with a normal hormone's actions. Thus, the raised concentration of human chorionic gonadotrophin during the first trimester of pregnancies complicated by hyperemesis has been implicated as a reason for the lower male:female ratio found in such pregnancies compared with normal ones (Sorensen 2000). Although these birth ratios may normalize with time, if the sex ratios are related to hormone levels at conception, they may one day be considered a biomarker of internal hormone status. A large prospective study measuring hormone levels and postnatal health (as well as gender) will be one way to test this hypothesis.
It is plausible that some proportion of the adverse health effects seen in some children could be the result of yet-to-be-recognized preimplantation exposures. Only by actually determining what's "on board" the mother around the time of conception and implantation can we identify those influences.

Maternal Nutritional Status
Maternal nutrition is thought to be an underlying determinant of a variety of pregnancy outcomes. A growing body of literature has established the association between maternal nutritional status and adverse pregnancy outcomes (Luke 1994a(Luke , 1994bRamakrishnan et al. 1999). Most studies have evaluated women well after the pregnancy has been established and the woman knows she is pregnant, typically during the second or third trimester. In contrast, the effect of a woman's nutritional status prior to pregnancy and during the first trimester on pregnancy outcomes has been rarely studied, even though evidence suggests that this is a vulnerable period in embryo and fetal development (Ashworth and Antipatis 2001;Czeizel 1995;Robinson et al. 1999). Studying women's nutritional status during this period will help determine the impact of early nutritional deficiencies or excesses on birth outcomes, and thereby assist in prevention efforts by enabling the design of more timely interventions.
Pregnancy is a time of increased cellular metabolism and rapid growth and development; therefore, there is an increased demand for energy and nutrients during pregnancy (King 2000). Changes in nutritional metabolism are driven by hormonal changes, fetal demands, and maternal nutritional supply. In general, well-nourished women need only a small amount of additional energy because of physiologic adaptations such as a lowered basal metabolic rate and reduced physical activity. However, these adaptations are limited in their capacity to adjust to changes in nutritional metabolism such that maternal undernutrition or micronutrient deficiencies or excesses during a critical stage of development may compromise fetal growth and development.
Nutritional requirements vary over the course of the pregnancy, depending on the stage of prenatal development. A woman's nutritional status prior to pregnancy, measured by prepregnancy weight, has been significantly associated with intrauterine growth retardation (weight for gestational age at birth < 10th percentile of a reference population) and low birth weight (< 2.5 kg) (WHO 1995). Higher maternal body mass index before pregnancy has been associated with a lower risk of having a small-for-gestational-age infant and an increased risk of late fetal death and NTDs (Cnattingius et al. 1998;Werler et al. 1996). Improving women's nutritional status in the months prior to pregnancy has resulted in higher birth weights and lengths (Caan et al. 1987). Data from the Collaborative Perinatal Project, a study of more than 55,000 pregnancies, found that maternal prepregnancy weight and gestational weight gain act independently of each other but are additive in their effect on fetal growth (Luke 1994b). It has been hypothesized that poor periconceptional nutrition may adversely affect placental growth, leading to insufficient fetoplacental exchange of nutrients, or that underlying micronutrient deficiency such as iron deficiency or anemia may stimulate the hypothalamic-pituitary-adrenal axis, thereby affecting early growth and length of gestation (Allen 2001;Robinson et al. 1995). A better understanding of the effects of preconceptional nutritional deficits (or excesses) and their causal mechanisms could lead to more effective interventions to reduce these fetal exposures and could help separate xenobiotic etiologies from nutritional causes of ill health.
Although nutrient requirements during early pregnancy are quantitatively small, specific nutritional deficiencies and toxicities can have adverse effects on prenatal development. During the first 2 weeks postovulation, maternal protein deficiency is associated with a variety of adverse pregnancy outcomes in animal studies (Desai et al. 1996;Gressens et al. 1997;Langley-Evans 2000). One such study found that offspring of female rats fed a low protein diet during preimplantation and then a normal control diet for the remainder of pregnancy had lower birth weights, increased growth rate postweaning, increased systolic blood pressure, and disproportionate growth of the liver and kidneys compared with control offspring (Kwong et al. 2000). Much of what we know about the importance of preimplantation nutrition in humans we have learned from human in vitro fertilization studies that suggest that certain amino acids (among other essential nutrients) are necessary substrates for the development and viability of the conceptus posttransfer (Conaghan et al. 1998;Devreker et al. 2001). Although evidence from human epidemiologic studies is not sufficient to conclude that there is a relationship between low protein intake in pregnancy and pregnancy outcomes (Metges 2001), studies of pregnant women exposed to extreme food deprivation suggest that this may be a vulnerable period in human development as well (Susser and Stein 1994). Specifically, women deprived during the 3-week period beginning after menstruation through implantation were more likely to have infertility problems and/or an infant born with an NTD. Those exposed in utero were more likely to have schizophrenia and schizoid and antisocial personality at 19 years of age.
During embryogenesis, the period of development from the third through eighth postovulatory weeks, an inappropriate dietary supply, a functional deficiency, or an excess amount of a micronutrient can be teratogenic. Although the causal mechanisms are not well understood in some cases, micronutrients are integral components of cellular metabolism. As such, micronutrients have the potential to impair cellular growth and replication in the rapidly growing embryo and fetus. Perhaps the best-known example of the impact of a micronutrient during this vulnerable period of development is the relation between folic acid deficiency and NTDs. In the last decade, periconceptional folic acid supplementation has been shown to prevent the occurrence and recurrence of NTDs (Lumley et al. 2002). NTDs occur within the first month after conception and are thought to be caused by failure of the neural tube to close (Sadler 1998). However, the mechanism by which folic acid prevents NTDs is unknown, and folic acid does not prevent all NTDs. Studies have suggested that the etiology of NTDs may be multifactorial, and deficiencies in other nutrients such as zinc, methionine, vitamin B12, and selenium may be involved (Guvenc et al. 1995;Kirke et al. 1993;Shoob et al. 2001;Velie et al. 1999). Excessive levels of micronutrients have also been associated with NTDs (e.g., vitamin A) (Rothman et al. 1995). In addition, there is increasing evidence that inadequate micronutrient status during organogenesis may be associated with an increased risk for other congenital anomalies such as orofacial clefts, conotruncal cardiac defects, and defects of the urinary system, as evidenced by reduced risk of these anomalies with periconceptional multivitamin supplementation (Botto et al. 1996;Czeizel 1996;Hall and Solehdin 1998;Itikala et al. 2001).
Micronutrient deficiencies have also been associated with adverse pregnancy outcomes such as preterm delivery (gestation < 37 weeks) and low birth weight. A recent review found an association between maternal anemia (i.e., low hemoglobin concentration) and these outcomes (Rasmussen 2001), but the majority of these studies document only the effect of second-trimester anemia. A few studies have found an association between firsttrimester anemia and preterm delivery (Scanlon et al. 2000;Scholl and Reilly 2000;Zhou et al. 1998). Although anemia is the most commonly used indicator for iron deficiency, iron deficiency is rarely assessed in pregnancy, and anemia is neither a sensitive nor specific indicator for iron deficiency. Some evidence suggests that iron deficiency anemia is associated with an increased risk for preterm delivery (Lu et al. 1991;Scholl et al. 1992), though this link is still unconfirmed (Rasmussen 2001). There is also evidence that deficiencies of other micronutrients such as zinc, calcium, and folic acid may be associated with preterm delivery and low birth weight.
Research into the area of periconceptional nutrition has shown that maternal nutritional deficiencies and toxicities in early pregnancy can have irreversible effects. Greater emphasis on research during the periconceptional period may yield data that confirm that other nutritional deficiencies or toxicities, previously unstudied in human populations, may affect health outcomes in the fetus, infant, and child. For a better understanding of the role of periconceptional nutrition on the health of the pregnant mother, fetus, and ultimately the infant, a woman's nutritional status should be evaluated during this critical time of development. Because women often do not know that they are pregnant until after the period of organogenesis, it is critical that any long-term study of health outcomes in children and adolescents enroll the woman before she becomes pregnant.
This review has focused on environmental exposures, although there are clearly other factors that were not considered, and which we know impact pregnancy outcome and the child's health, including a large number of diseases (reviewed in Geist and Koren 2001) and environmental heat (Wells and Cole 2002).

Conclusions
Even this modest data review shows that we know a considerable amount about the relatively short-term consequences of various situations that occur before and at the time of conception. Studies have shown that numerous xenobiotics are present in semen and follicular fluid, or circulating in maternal blood and lodged in fat stores. The expanding ART literature clearly documents that the environment at the time of conception can impact the developing embryo, and thus likely affects the long-term health of the child, and thus affects healthcare spending and its associated burdens. Nutrition clearly impacts the progress of the pregnancy and the subsequent health of the child in many known ways.
This review documents what is already known about the determinants of children's health. Knowing that we only make new findings (i.e., become smarter) when we ask new questions, the implication that cannot be ignored is that the long-term longitudinal National Children's Study (2002) is our best hope for actually testing some of the potential effects identified in this literature, and for identifying new associations between early exposures and later health status. Only by collecting biological specimens and actually measuring what the internal exposures are in women and men prior to and at the time of conception will we be able to definitively say which, if any, of these exposures are meaningful. Doing this in the context of a large longitudinal epidemiologic health study will help us determine which of these exposures impacts the health of the offspring. Knowing this can powerfully inform future decision-making, will create hypotheses to test and verify for years to come, and will allow regulatory efforts to focus on those exposures that have the greatest impact. This is a chance that must not be missed.