Evaluation of the Association between Maternal Smoking, Childhood Obesity, and Metabolic Disorders: A National Toxicology Program Workshop Review

A PubMed (http://www.ncbi.nlm.nih.gov/pubmed) search strategy updated through 8 March 2012 was developed to identify human and animal studies related to maternal smoking or tobacco exposure during pregnancy and a) weight, growth, or adiposity after infancy (≥ 1 year of age in humans, after postnatal day 21 in rodents), or b) health outcomes in offspring at any age related to T1D or T2D, glycemic control, or metabolic syndrome. Studies assessing growth in infants < 1 year of age were excluded because maternal smoking during pregnancy is a known risk factor for low birth weight, whereas the primary question being addressed at the workshop was whether maternal smoking is a risk factor for childhood obesity. The literature search included both MeSH-based and key word strategies to identify articles not yet indexed in PubMed [see Supplemental Material for a complete list of search terms (https://doi.org/10.1289/ehp.1205404)]. This search identified a total of 1,551 publications (see Supplemental Material, Figure S1). Of these, 99 studies presented original data that met our inclusion criteria and were considered relevant as the primary literature. Two additional epidemiological studies identified during the peer-review process were also included (Fasting et al. 2009; Mendez et al. 2008) for a total of 101 studies (83 human studies and 18 animal studies).

Offspring overweight or obesity. Smoking during pregnancy is a known risk factor for low birth weight and small-for-gestational-age infants (Lumley et al. 2008; Oken et al. 2008; Samper et al. 2011), but is increasingly accepted as a risk factor for childhood overweight and obesity based on the consistent positive associations reported among studies (Figure 1). Most studies (34 of 42) summarized in Figure 1 present results that support a causal link between maternal smoking and subsequent child overweight or obesity. Many of these studies were evaluated in two recent meta-analyses (Ino 2010; Oken et al. 2008). Based on results reported in 14 observational studies, maternal smoking during pregnancy was associated with overweight at 3–33 years of age [pooled adjusted odds ratio (adjOR) = 1.50; 95% CI: 1.36, 1.65] (Oken et al. 2008). The pooled OR estimated by Ino (2010) for obesity [body mass index (BMI) > 95th percentile] was 1.64 (95% CI: 1.42, 1.90) based on 16 studies. Women who smoked during pregnancy tended to weigh more, and tended to have lower socioeconomic status, have less education, and were less likely to breastfeed (Ino 2010; Oken et al. 2008). Both meta-analyses used funnel plot methods to ascertain publication bias and concluded there was some evidence for publication bias, but not enough to negate the overall conclusion of increased risk.

Ethnicity may affect the association between maternal smoking and childhood obesity. A large retrospective study by Sharma et al. (2008) of the PedNSS (Pediatric Nutrition Surveillance System) cohort found significant associations between maternal smoking and childhood obesity only in white (n = 66,191) and black (n = 28,230) children; the associations were not apparent for Hispanic (n = 34,378), American Indian/Alaska native (n = 2,228), or Asian/Pacific Islander children (n = 4,740). The lack of association in children of Asian/Pacific Island descent is interesting because studies on Japanese children have reported findings consistent with a positive relationship (Ino et al. 2011; Mizutani et al. 2007; Suzuki et al. 2009, 2011). Too few studies are available on Native American and Hispanic children to interpret the findings of Sharma et al. (2008) in the context of other studies. One Brazilian study by Tome et al. (2007) did not find an association between maternal smoking and offspring overweight at 8–10 years of age (adjOR = 1.07; 95% CI: 0.84, 1.37); however, they did find that children of mothers who smoked were less likely to be underweight (adjOR = 0.56; 95% CI: 0.40, 0.78). It is worth noting that this study was conducted in the Ribeirao region of Brazil where 9.5% of the children were malnourished (BMI < 5th percentile).

Unmeasured confounding. Although the literature on the association between maternal smoking and obesity/overweight in children is fairly consistent when looking across studies, there are some indications that unmeasured familial or lifestyle factors may contribute to the association. One study conducted sibling-based analyses to evaluate the relationship between maternal smoking and BMI in 124,203 males born between 1983 and 1988, identified in the Swedish Medical Birth Registry with the intent of investigating the impact of familial factors on the association between maternal smoking and child obesity (Iliadou et al. 2010). They reported an association between maternal smoking during the first trimester and overweight (BMI ≥ 25) for the cohort as a whole (adjOR = 1.56; 95% CI: 1.46, 1.66 for > 10 cigarettes/day compared with 0 cigarettes). However, the authors found smaller associations based on within-family analyses, suggesting partial confounding by familial factors. Gilman et al. (2008) also conducted a sibling-based analysis of children enrolled in the Collaborative Perinatal Project to address the issue of unmeasured confounding factors by fitting conditional fixed-effects models among siblings. This type of analysis capitalizes on variability in exposure to tobacco smoke that can occur among siblings when the mother’s smoking pattern during gestation differed across pregnancies. Conceptually, the sibling analysis matches exposed and unexposed siblings on family background to identify evidence of unmeasured sources of confounding. Gilman et al. (2008) found significant positive associations in the full cohort for overweight at 7 years of age were still apparent in the conditional sibling fixed effects analysis.

Secondhand smoke exposure during pregnancy or childhood. Although the literature search strategy and inclusion criteria were constructed to identify studies assessing maternal smoking during pregnancy, several studies of paternal or “partner” smoking during pregnancy (Durmus et al. 2011b; Kwok et al. 2010; Leary et al. 2006; Matijasevich et al. 2011) or during childhood (von Kries et al. 2008) were identified. Some of these studies reported no association between parental/partner smoking and BMI or adiposity end points (Durmus et al. 2011b; Matijasevich et al. 2011) whereas others did (Kwok et al. 2010; Leary et al. 2006; von Kries et al. 2008), suggesting that paternal/partner smoking may be a contributing factor on its own, either by unmeasured familial factors, or as a contributor to maternal exposure from secondhand smoke. Another study by Steur et al. (2010) assessed predictors of obesity at 8 years of age in 1,687 children participating in the Prevention and Incidence of Asthma and Mite Allergy (PIAMA) study, a prospective study based in the Netherlands. Maternal smoking during pregnancy and “smoking in the parental house” were both included in the model selection, but only “smoking in the parental house” was selected as a significant independent predictor of childhood overweight.

Other measures of adiposity, BMI, and leptin. The findings discussed above focus on studies of associations between maternal smoking and overweight or obesity. Other studies have estimated associations between maternal smoking during pregnancy and continuous measures of BMI, standardized BMI (i.e., z-score or SD score), or body weight in the offspring [see Supplemental Material, Table S1 (https://doi.org/10.1289/ehp.1205404)] (Beyerlein et al. 2010, 2011; Durmus et al. 2011a; Goldani et al. 2007; Hill et al. 2005; Iliadou et al. 2010; Kanellopoulos et al. 2007; Leary et al. 2006; Matijasevich et al. 2011; Ravnborg et al. 2011; Suzuki et al. 2011; Syme et al. 2010). Overall, these studies have reported positive associations with maternal smoking during pregnancy and child BMI. There are also indications that the children of mothers who smoked during pregnancy display more rapid weight gain earlier in life (Griffiths et al. 2010; Karaolis-Danckert et al. 2008; Wen et al. 2011).

Studies in the literature often assess obesity based on BMI, which is not necessarily a direct measure of adiposity because it does not distinguish fat and lean mass (Freedman and Sherry 2009; Wells 2001). Associations between maternal smoking and indicators of adiposity (i.e., body fat, abdominal fat, skin fold thickness) [see Supplemental Material, Table S2 (https://doi.org/10.1289/ehp.1205404); Durmus et al. 2011a; Leary et al. 2006; Silva et al. 2010; Syme et al. 2010; Vik et al. 1996; von Schnurbein et al. 2011] are less consistent than associations with overweight/obesity or continuous measures of BMI-related end points (Figure 1). Given the limitations of BMI as a measure of adiposity, it is therefore important for future studies to consider other more direct measures of adiposity/fat mass to clarify the contribution of maternal smoking to childhood obesity.

Leptin, a protein produced by adipocytes and placental tissue, appears to function as a link between adiposity, satiety, and energy regulation (Lee and Fried 2009; Pelleymounter et al. 1995; Woods and D’Alessio 2008). Leptin levels have been correlated with neonatal growth (Christou et al. 2001; Vatten et al. 2002), but very few studies have looked at leptin levels beyond the neonatal period, and it is unclear how informative the infant studies are for addressing issues related to childhood obesity, given the complex role that maternal smoking has on childhood growth (i.e., as a risk factor for low birth weight and for childhood obesity). Overall, the findings on leptin levels in cord blood or blood collected from infants in association with maternal smoking status are mixed, with some studies reporting positive associations (Helland et al. 2001), negative associations (Mantzoros et al. 1997; Ozkan et al. 2005), or no association (Coutant et al. 2001; Kayemba-Kay’s et al. 2008; Pardo et al. 2004; Vatten et al. 2002).

T1D. The existing human studies do not provide support for an association between maternal or paternal smoking during pregnancy and childhood T1D (Figure 2). Only 2 of the 13 studies assessing end points related to T1D following exposure during pregnancy reported findings consistent with an association (Sipetic et al. 2005; Wahlberg et al. 2011). In the study by Sipetic et al. (2005), the frequency of maternal smoking during pregnancy was higher in T1D cases compared with controls (37.2% vs. 24.8%, p = 0.023 or unadjusted OR = 1.79; 95% CI: 1.08, 2.97). Although the univariate finding was statistically significant, the authors did not include maternal smoking in the multivariate analysis conducted to identify risk factors for T1D. In the study by Wahlberg et al. (2011), a positive association was observed for one indicator of β-cell autoimmunity in T1D, tyrosine phosphatase (IA-2A) + antibody status (unadjusted OR = 1.6; 95% CI: 1.2, 2.2), and not for glutamic acid decarboxylase (GAD) + antibody status (data not shown). None of the five prospective studies reported a positive association between maternal or paternal smoking during pregnancy and T1D (Hummel et al. 2001; Johansson et al. 2008; Rosenbauer et al. 2007; Stene et al. 2004; Toschke et al. 2007a; Wadsworth et al. 1997), including two studies of children who were considered at high risk for developing the disease (Hummel et al. 2001; Stene et al. 2004). Several studies reported that T1D was less likely among children whose mothers smoked during pregnancy than among the children of nonsmoking mothers (Dahlquist and Kallen 1992; Marshall et al. 2004; Svensson et al. 2005) for T1D. Hummel et al. (2001) reported a nonsignificant negative association between islet-cell antibodies (a marker of β-cell autoimmunity) and maternal smoking during pregnancy among children considered at high risk because one or both parents had T1D. However, the number of cases in this study was small (47 of 1,043 children). One hypothesis that may explain evidence of a protective effect of exposure to smoking relates to the hygiene theory of autoimmune disease; this suggests that the risk of autoimmune disease may be increased if the environment is “cleaner,” which might be manifest in the households of nonsmokers (Johansson et al. 2008; Marshall et al. 2004). Alternatively, smoking may affect immune function, resulting in a reduced risk of developing autoimmune disorder (Prescott 2008).

Although the overall literature does not support an association between maternal smoking during pregnancy and T1D in offspring, associations with exposure to smoke during childhood should be investigated further. Three studies have assessed childhood exposure and T1D, and both reported findings supportive of an association (Hathout et al. 2006; Sipetic et al. 2005; Skrodeniene et al. 2008). In the study by Skrodeniene et al. (2008), Lithuanian children who were islet-cell autoantibody positive were more likely to live in a home where family members smoked indoors than children who were islet-cell autoantibody negative [7 of 13 (53.8%) vs. 53 of 199 (26.6%), respectively, p = 0.004; unadjusted OR = 3.21; 95% CI: 1.03, 10.00]. The frequency of mothers who reported smoking during pregnancy did not differ between the groups in this study. Hathout et al. (2006) reported that children with T1D were more likely to have been exposed to passive smoking than control children in Southern California [31 of 102 cases (30%) vs. 3 of 10 controls (10%), p = 0.0001; unadjusted OR = 3.90; 95% CI: 1.19, 17.24]. One limitation in the existing literature that needs to be addressed in future studies investigating the association of smoking or other environmental exposures and T1D is use of statistical models that consider potential confounding factors (e.g., viruses, nutrition, or socioeconomic psychosocial factors).

T2D and metabolic syndrome. Fewer studies have looked at associations between maternal smoking during pregnancy and T2D, blood glucose, blood insulin, or metabolic disease (Horta et al. 2011; Huang et al. 2007; Montgomery and Ekbom 2002; Thiering et al. 2011; Thomas et al. 2007) (Figure 3). It is not possible to reach conclusions with confidence based on the existing literature because of the variation in health outcomes assessed among the studies [i.e., HbA1C (hemoglobin A1C); fasting glucose, nonfasting glucose, HOMA-IR (homeostatic model assessment–insulin resistance), blood insulin; T2D; and metabolic syndrome. In addition, although most of the studies do not report positive associations with these outcomes, the age at assessment in many of the studies was relatively young, and additional follow-up may be required. It is worth noting that no association with HbA1C ≥ 6% was found in the oldest and largest cohort assessed [7,518 men and women from the 1958 British birth cohort enrolled in the Perinatal Mortality Survey (PMS) evaluated at 45 years of age] (Thomas et al. 2007). This finding does not confirm an earlier report for the same cohort of an association between heavy smoking during pregnancy and T2D in offspring at 33 years of age (adjOR = 4.02; 95% CI: 1.14, 14.14) (Montgomery and Ekbom 2002). A related analysis of the cohort by Power et al. (2010) also reported no association with HbA1C in adults at 45 years of age (fully adjusted mean difference = 0.13; 95% CI: –0.33, 0.58). Power et al. (2010) reported a positive association between maternal smoking during pregnancy and metabolic syndrome based on an unadjusted model [unadjusted OR = 1.21; 95% CI: 1.05, 1.39), but there was a significant negative association between maternal smoking and metabolic syndrome after adjustment for adult life covariates such as social class, education, physical activity and inactivity (television/computer use), smoking, and consumption of fruit/vegetables, cakes/sweets, and alcohol (adjOR = 0.55; 95% CI: 0.47, 0.64).

There are 599 known cigarette additives (Rabinoff et al. 2007), and most are uncharacterized for potential toxicity, including metabolic effects. One exception is cadmium, a constituent of tobacco of smoke shown to alter glucose homeostasis or insulin sensitivity in animals exposed as adults (Edwards and Prozialeck 2009) but has not been assessed for metabolic effects after developmental exposure. Only two studies assessed prenatal exposure to cigarette smoke during pregnancy and body weight later in life in offspring (Chen et al. 2011; Ng et al. 2009).

Body weight and adiposity. Most animal studies have reported that rats exposed to nicotine during perinatal development tended to have a higher body weight and more fat mass compared with controls (Figure 4), with the effect typically first apparent at weaning and persisting through adulthood (Gao et al. 2005; Newman et al. 1999; Oliveira et al. 2009, 2010; Somm et al. 2008). However, there are exceptions to this pattern (Gao et al. 2008; Holloway et al. 2007), which do not appear to be attributable to strain differences, dosing, or the timing of exposure. Food intake was unaffected in studies that evaluated it as an end point (Oliveira et al. 2009, 2010; Somm et al. 2008). The potential interaction between perinatal nicotine exposure and postnatal diet has been explored in only one study, which examined whether a high-fat diet (HFD) postnatally would exacerbate the weight gain in nicotine-exposed animals (Somm et al. 2008). Although the nicotine-treated animals that were fed an HFD did not consume more kilocalories than controls fed the same diet, they were heavier. The nicotine-treated animals were less physically active than control animals that consumed the same diet. There were no differences in oxygen consumption or respiratory exchange ratio, suggesting that increased weight gain in HFD animals with the same caloric intake was perhaps attributable to the dysregulation of adipose tissue development, thereby leading to higher amounts of fat storage.

Glucose homeostasis and insulin sensitivity. There are several reports in the literature on impaired glucose homeostasis in the male offspring of rats that were treated with nicotine during gestation (Somm et al. 2008), lactation (Oliveira et al. 2010), or gestation and lactation (Bruin et al. 2007, 2008c; Holloway et al. 2005) (Figure 5). Administered doses ranged from 1 to 6 mg/kg-day delivered to the mother either via an osmotic minipump implanted under the skin or via daily subcutaneous injections. These dosing protocols resulted in maternal serum cotinine levels considered relevant to women who smoke or use nicotine patches as cigarette substitutes (Bruin et al. 2007; Hackman et al. 1999; Somm et al. 2009). The most consistent findings from these studies were indications of insulin resistance in adulthood based on increased insulin area under the curve (AUC) following an oral or intraperitoneal glucose challenge (Bruin et al. 2007, 2008c; Holloway et al. 2005; Somm et al. 2008) or on an increased insulin resistance index (Oliveira et al. 2010). These effects were seen across studies despite the use of different doses of nicotine, different administration protocols, and different windows of developmental exposure within the perinatal period (i.e., gestation and/or lactation). A study by Holloway et al. (2007), reported increased insulin resistance in 15-week-old F2 generation male offspring of dams that were treated with nicotine only during gestation and lactation compared with untreated controls. Impairments in glucose tolerance also have been reported following nicotine exposure (Bruin et al. 2007, 2008c; Holloway et al. 2005; Somm et al. 2008). Effects on fasting insulin levels have been less consistent, with some studies showing no effect (Bruin et al. 2007; Somm et al. 2008) and others reporting increased levels in adult rodents compared with untreated controls (Holloway et al. 2005, 2007; Oliveira et al. 2010). At low dose levels (1 mg/kg-day) permanent changes in glucose homeostasis were observed only when exposure occurred during both fetal and neonatal life (i.e., pregnancy and lactation); however, when dams were exposed to 3 mg/kg-day (Somm et al. 2008) or 6 mg/kg-day (Oliveira et al. 2010), then either fetal or lactational exposure was sufficient to affect glycemic control. Importantly, no effects on glucose homeostasis were observed in Wistar rats (Jose et al. 2009) or in a series of studies (Jose et al. 2009; Swislocki 2003; Swislocki and Fakiri 2008; Swislocki et al. 1997) in Sprague-Dawley rats treated with nicotine after weaning. Taken together, these data suggest that cigarette smoke or nicotine exposure during gestation and lactation is the critical window of exposure.

Pancreatic effects. Changes in pancreatic weight, morphology, and function have been reported in animals that were treated with nicotine during gestation alone or during gestation and lactation (Bruin et al. 2007, 2008a, 2008b; Grove et al. 2001; Holloway et al. 2005; Somm et al. 2008) (Figure 6). These pancreatic effects include increased β-cell apoptosis and decreased β-cell mass in rodents (Bruin et al. 2007; Holloway et al. 2005; Somm et al. 2009). Other findings include decreased pancreatic weight in infant rhesus monkeys (Grove et al. 2001). As noted earlier, the treatment protocols used in the rodent studies result in maternal serum cotinine levels that are considered relevant to women who smoke or use nicotine patches as cigarette substitutes. The pancreatic effects are hypothesized to occur as a direct effect of nicotine binding to nicotinic acetylcholine receptors in the developing pancreas, leading to oxidative stress and mitochondrial damage and eventual β-cell apoptosis (Bruin et al. 2007, 2008a, 2008b, 2008c). Pancreatic tissue is considered to be especially susceptible to oxidative stress–mediated tissue damage due to low levels of antioxidant enzyme expression (Lenzen et al. 1996; Tiedge et al. 1997).

Leptin signaling. Adipose tissue is increasingly recognized as an active endocrine organ with many secretory products. In addition, adipose tissue is considered to be part of the innate immune system (Gaspari et al. 2011). Leptin is one of the key hormones involved in feeding behavior and energy expenditure [reviewed by Belgardt et al. (2010)]. In general terms, leptin is secreted from the adipocytes of white adipose tissue and delivers information on peripheral energy stores to the central nervous system. Leptin resistance that involves impairment in leptin transport, leptin signaling, and/or the neurocircuitry of energy balance is a risk factor for obesity (Morris and Rui 2009). Several studies have assessed leptin levels or the expression of proteins involved in leptin signaling after developmental nicotine exposure (Grove et al. 2001; Huang and Winzer-Serhan 2007; Santos-Silva et al. 2010). Findings from animal studies reveal that early exposure to nicotine appears to be associated with dysregulation of leptin levels, likely via hypertrophy of adipocytes and increased gene expression of pro-adipogenic transcription factors (Somm et al. 2008) and hypothalamic leptin signaling (Oliveira et al. 2009, 2010).

The association between smoking and leptin levels in human infants remains controversial, with both hyper- and hypoleptinemia being reported (Coutant et al. 2001; Helland et al. 2001; Kayemba-Kay’s et al. 2008; Mantzoros et al. 1997; Ozkan et al. 2005; Pardo et al. 2004, 2005; Vatten et al. 2002).

Based on human epidemiological studies and animal experiments reported in the literature, current evidence supports a causal relationship between maternal smoking and increased risk of obesity or overweight in offspring; however, the possibility that the association is attributable to unmeasured residual confounding cannot be completely ruled out. The literature is not sufficient to evaluate the potential impact of maternal smoking after delivery, or the contribution of exposure to secondhand smoke during pregnancy and early childhood. Existing studies in humans do not provide strong support for an association between maternal or parental smoking and childhood T1D, and too few epidemiological studies have assessed outcomes related to T2D to reach conclusions. However, findings from animal studies identify associations between perinatal nicotine exposure and disruption of pathways important in the pathophysiology of T2D, including reduced β-cell mass and impaired β-cell function.

In Appendix 1 we identify research gaps and provide specific study recommendations noting in particular the need to better understand the a) relationship between developmental exposure to tobacco smoke as a risk factor for T2D or metabolic syndrome, b) consequences of pre- and postnatal exposure to secondhand smoke (also called environmental tobacco smoke, involuntary smoke, and passive smoke), c) impact of timing of exposure, dose, duration and route (e.g., nicotine replacement therapy), and d) mechanistic basis for the association between cigarette smoke or nicotine exposure with childhood obesity and metabolic outcomes. Elucidating the mechanistic basis for these associations could be an important component of developing a toxicological screening approach that could be used to identify other environmental chemicals for targeted assessment of obesogenic and metabolic effects. For example, a number of pesticides included in Phase 1 of the U.S. Environmental Protection Agency (U.S. EPA 2012) ToxCast™ high throughput screening program interact with nicotinic receptors but have never been assessed for potential adipogenic or metabolic effects [see Supplemental Material, Figure S2 (https://doi.org/10.1289/ehp.1205404)]. Addressing the data gaps can help us understand more about the nature of the association as well as inform decisions regarding intervention strategies. Epidemiological studies will also need to consider the challenges related to validity of self-reported maternal tobacco use or secondhand smoke exposure, especially in studies evaluating the long-term health consequences for offspring where exposure assessment may occur many years after the exposure event. With respect to mechanistic research, more attention should be given to the possibility that environmental exposures could alter central nervous system regulation of energy and body weight homeostasis. The brain is the primary organ involved with determining behavior and obesity; metabolic diseases are often considered to be manifestations of behavioral excess (i.e., consuming calories in excess of one’s energy needs). A number of pathways are important for the development of obesity and/or diabetes, including but not limited to alterations in neurohumoral signaling, feeding behavior, energy balance, brain and peripheral inflammation, and insulin resistance. To date, the effects of maternal cigarette smoking and/or nicotine exposure on these pathways remains largely unexplored (Wilborn et al. 2005; Woods 2009; Woods and D’Alessio 2008; Zheng et al. 2009). Another largely unexplored hypothesis is that fetal/neonatal exposure to smoking/nicotine sensitizes the offspring to adverse effects of obesogenic diets. Although perturbations in the microbiome profile have been associated with diabetes and obesity, the contribution of or interactions with smoking have not been explored and may also influence susceptibility to adverse diets.

The association between maternal smoking and childhood obesity adds to the recognized health burden from tobacco exposure, which is estimated at almost 5 million deaths per year globally (World Health Organization 2010). Maternal smoking during pregnancy may not account for as many cases of childhood obesity or overweight as more traditional risk factors such as decreased physical activity, watching television > 1 hr/day, and low parental education level (Toschke et al. 2007b), but it is a risk factor that, if prevented, can help reduce the burden of childhood obesity at the population level. From a toxicological perspective, the linkages between maternal smoking during pregnancy and childhood overweight/obesity provide proof-of-concept of how early-life exposure to an environmental toxicant can act as a risk factor for childhood obesity.

Epidemiology

––RT during pregnancy trials, NRT following smoking cessation, nicotine therapeutics for nonsmokers, Snus (a moist powder tobacco product similar to dry snuff)

––Other data may be available from FDA for nicotinic acetylcholine receptor agonist drugs, and putative “reduced” harm cigarettes

Animal and in vitro

We gratefully acknowledge the contributions of S. Holmgren (NIEHS) for developing the literature search strategy, and J. Stevens (GLP Support Services) and K. Taylor (NTP/NIEHS) for assistance in preparing the background literature review document. We also acknowledge the invaluable assistance of the staff at the U.S. EPA National Center for Computation Toxicology (NCCT), particularly D. Reif, D. Dix, K. Houck, and R. Judson. We also thank J. Heindel and M. Longnecker (NIEHS) for their excellent review of the manuscript.