Developmental Programming

Developmental programming, including the possible role of epigenetics, as a potential determinant of GWG, is discussed in Chapter 4. In this chapter, the role of developmental programming as a mechanism for some of the effects of GWG on postnatal outcomes is discussed. Many animal models have demonstrated that altering the environment in utero can have lifelong consequences. Perturbations of the maternal diet during pregnancy (typically by severe energy or protein restriction; administration of hormones such as glucocorticoids; mechanical means, such as ligation of the uterine artery; or induction of anemia or hypoxia) have postnatal consequences on a number of metabolic and behavioral traits. Effects are inducible in rodents and other mammals, including non-human primates. Inasmuch as humans differ from other animal species in duration of pregnancy, placentation, and other important factors, the importance of the findings from animal studies lies not in the specific interventions but rather in the general principle that altering the supply of nutrients, hormones, and oxygen to the growing embryo and fetus or exposing them to stressors and toxicants can have long-term effects. Much of this animal research has focused on obesity-related outcomes such as adiposity, fat distribution, sarcopenia, insulin sensitivity, glucose intolerance, and blood pressure. These are related to the leading causes of morbidity—and ultimately, mortality—in the United States. The ways in which GWG could influence obesity-related child health outcomes through developmental programming is discussed below (in Childhood Obesity and Its Consequences).

Until recently, most of the research in animal models concentrated on the long-term effects of interventions that cause offspring to be born small, typically small-for-gestational age (SGA), rather than early. Such work has been a good companion to a series of epidemiologic observations made within the past two decades that lower birth weight, apparently resulting from both reduced fetal growth and reduced length of gestation, is associated with higher risks of central obesity, insulin resistance, the metabolic syndrome, type 2 diabetes, hypertension, and coronary heart disease later in life. These associations are potentiated by rapid weight gain in childhood (Bhargava et al., 2004; Barker et al., 2005).

It is important to note, however, that in recent years researchers have recognized that higher birth weight is also associated with later obesity and its consequences. Greater GWG is associated with increased weight at birth (reviewed in the Fetal Growth section of this chapter) based on either the absolute amount of GWG and indicators of excessive gain (based on total GWG relative to the recommendations for gain within a given prepregnancy BMI category). Excessive GWG appears to be rising over time (see Chapter 2), highlighting questions about the long-term adverse effects of higher weight gains in pregnancy. Animal experiments that involve “over-nutrition” of the mother during pregnancy are discussed briefly below.

In addition, it is critical to recognize that effects of GWG, or indeed any factor that alters the in utero environment, may have long-term effects on the offspring without any alterations of fetal growth or length of gestation. Thus the most important epidemiologic evidence for long-term effects of GWG does not depend on birth weight, gestational age, or birth weight for gestational age as exposures or outcomes, but rather provides data on the direct associations of GWG with various health outcomes in the offspring. With this in mind, the committee considered “fetal growth” outcomes, including SGA and large-for-gestational age (LGA), and preterm birth as short-term outcomes. These measures have demonstrable and substantial associations with neonatal morbidity and mortality. Other short-term outcomes include stillbirth and birth defects. In contrast, neonatal body composition is included in the discussion of long-term outcomes because of the hypothesis (still unproven, however) that relative amounts of adiposity and lean mass—and their physiologic consequences—in fetal and neonatal life are important in setting long-term cardio-metabolic trajectories.

It also bears noting that this report focuses primarily on GWG, rather than prepregnancy BMI. Nevertheless, because the two factors are closely linked, one must account for confounding and effect modification by BMI in addressing offspring effects of GWG. Also it is possible that factors in infancy or childhood (e.g., growth in stature, adiposity, and infant feeding) could mediate effects of GWG on long-term child health.

Childhood Obesity and Its Consequences

The following discussion focuses primarily on mechanisms linking GWG to childhood obesity and its consequences, although similar mechanisms likely underlie associations of GWG with fetal growth. One issue that hampers inferences regarding fetal growth is that fetal growth is usually characterized by (gestational-age-specific) weight at birth, with less consideration of trajectory from the time of conception to delivery of weight, body length, or body composition (see Chapters 3 and 4 for a review of existing studies that address these issues). In contrast to the prenatal period, serial measurements of length/height and weight are common during childhood but data on body composition are relatively scarce.

Insulin resistance and glucose intolerance during pregnancy may mediate effects of GWG on long-term child outcomes. Weight gain in pregnancy is partly a gain in adiposity, which is accompanied by a state of relative insulin resistance starting in mid-pregnancy, among other metabolic alterations (Reece et al., 1994; Williams, 2003; Catalano et al., 2006; King, 2006; Hwang et al., 2007) (also see Chapter 3). This is an adaptive response, as it allows more efficient transfer of fuels across the placenta to the growing fetus (King, 2006). In overweight and obese pregnant women, these changes are magnified; insulin resistance is more severe than in normal weight women, substantially raising the risk of impaired glucose tolerance and frank gestational diabetes mellitus.

This increased risk of impaired glucose tolerance has consequences for the fetus since glucose freely crosses the placenta; specifically, in pregnant women who have hyperglycemia, the fetus also experiences hyperglycemia. In a hypothesized sequence that Freinkel et al. (1986) termed “fuel-mediated teratogenesis,” fetal hyperglycemia causes fetal hyperinsulinemia, which in turn causes increased adiposity in the fetus. This increase is manifest as larger size at birth, which translates into higher rates of LGA and lower rates of SGA newborns (see discussion below and in Chapter 3). Presumably through developmental programming mechanisms, increased fetal adiposity also results in increased adiposity in the growing child. Other fuels besides glucose may also be involved. For example, increased fetal production of anabolic hormones and growth factors, in combination with the increased levels of glucose, lipids, and amino acids that are typical of GDM, can cause fetal macrosomia (birth weight > 4,500 g) and increase the risk for neonatal complications (Catalano et al., 2003). Crowther et al. (2005) and Pirc et al. (2007) showed that diet and insulin therapy along with blood glucose monitoring in pregnant women with mild GDM could lower plasma insulin and leptin (but not glucose) concentrations in cord blood, decreasing the risk of macrosomia by more than 50 percent (Crowther et al., 2005).

This same impaired physiologic milieu may also increase the risk for long-term complications, particularly obesity and its metabolic sequelae. Observational studies suggest that this may be the case. For example, among 5- to 7-year-old children in two American health plans, Hillier et al. (2007) showed that risk of high weight for age was lower among those whose mothers had been treated for GDM than those who had not been treated; the weight status of the “treated” offspring was similar to those whose mothers had normal glucose tolerance. However, long-term child follow-up studies and relevant randomized trials are necessary to conclusively determine if treatment of GDM or impaired glucose intolerance during pregnancy can reduce adiposity and related physiology.

Most of the evidence in support of the Freinkel hypothesis comes from animal experiments, such as those of van Assche and colleagues (1979), and more recently Plagemann and colleagues (1998). By pharmacologically induced GDM in rats, both groups of researchers observed fetal hyperglycemia and hyperinsulinemia, as hypothesized, as well as changes in the hypothalamus that give rise to hyperphagia, overweight, and impaired glucose tolerance in maturing offspring. Another way to induce offspring metabolic derangement in rats is through overfeeding the pregnant dam. For example, Samuelsson et al. (2008) reported that maternal diet-induced obesity resulted in increased adult adiposity and evidence of cardiovascular and metabolic dysfunction in the offspring (which was not present in the offspring of lean dams). Earlier work by Dorner et al. (1988) and Diaz and Taylor (1998) showed that a period of overfeeding or GDM in the pregnant dam during a developmentally sensitive period in gestation not only could change the metabolic phenotype of the immediate offspring, but also that the induced metabolic phenotype persisted for two succeeding generations. In their review of animal studies, Aerts and Van Assche (2003) demonstrate that these intergenerational physiologic effects are maternally transmitted, most likely through epigenetic processes. Seemingly paradoxically, in animal experiments it is also possible to produce offspring that have insulin resistance, features of the metabolic syndrome, and diabetes, including GDM, by reducing energy or macronutrient intake of the mother during pregnancy. This situation can also result in intergenerational amplification of obesity and its consequences. For example, in rats, Benyshek et al. (2006) were able to alter glucose metabolism in the grand-offspring by restricting protein during pregnancy and lactation.

In summary, animal experiments show that offspring obesity and related metabolic sequelae can be induced experimentally, either through pharmacological induction of GDM or through either over- or underfeeding pregnant dams as well as through mechanical means like uterine artery ligation. Epigenetic modifications likely explain many of these phenomena (Simmons, 2007). A human counterpart to the animal experimental work is epidemiologic studies showing that higher birth weight is related to later obesity and type 2 diabetes while lower birth weight is associated with central obesity, the metabolic syndrome, and indeed, type 2 diabetes as well (Gillman, 2005). In other words, a U-shape relationship exists between birth weight and obesity-related health outcomes.

The extent to which these observations on metabolic dysfunction and offspring obesity have relevance for GWG guidelines is still unclear. Few animal studies directly assess the influence of GWG on short- or long-term offspring outcomes. Animal experimentalists typically do not measure weight gain during pregnancy, and it is not clear whether appropriate animal models exist to study GWG and offspring obesity-related outcomes. Neither is it clear that models of either diet-induced obesity or GDM are instructive for assessing effects of GWG.

Likewise, human population studies that rely on birth weight or its components, duration of gestation, and size at birth as predictors of later outcomes (e.g., Hofman et al., 2004; Hovi et al., 2007) also do not directly assess GWG. Further, intervention studies to treat GDM do not in themselves provide evidence for making recommendations for appropriate GWG. Only randomized trials that alter weight gain during pregnancy can address that goal directly. In a randomized controlled trial of reduced weight gain among obese pregnant women, Wolff and colleagues (2008) reported that reduced weight gain led to reduced insulin and leptin concentrations but that glucose values were hardly altered. Mean weight gain in the intervention group was 6.6 kg (± 5.5 kg) vs. 13.3 kg (± 7.5 kg) in the control group a mean difference of 6.7 kg (95% CI: 2.6–10.8, p = 0.002). Although the study was small, with only 50 participants, the results nonetheless raise the possibility that moderating GWG may reduce the risk of GDM and, in turn, childhood obesity, but larger and longer-term studies are needed to address this question directly.

AHRQ Review of Studies on the Association Between GWG and Birth Weight

The Agency for Healthcare Research and Quality (AHRQ) evidence-based review on outcomes related to GWG (Viswanathan et al., 2008) identified 25 studies of variable quality that examined GWG and birth weight as a continuous measure. Every one of those studies demonstrated an association between higher GWG and higher infant birth weight. Although there was substantial variability in magnitude of effect across studies, in general birth weight differed by about 300 g between the lowest and highest GWG categories. Among the stronger studies, the AHRQ review found that for each 1 kg increase in GWG, birth weight rose 16.7–22.6 g. The fewer studies that considered weight gain by trimester tended to show a smaller increase in birth weight per unit increase in GWG in the third than in the first or second trimesters.

A smaller but still sizable number of studies (13) examined the relationship of GWG to risk of low birth weight (LBW, defined as < 2,500 g). These studies showed that risk of LBW diminishes as GWG increases, particularly as total gain exceeds 25–30 pounds. Although the magnitude of association varied substantially across studies, in general the highest GWG category had roughly half the risk of an LBW infant compared to the lowest GWG category. At the other end of the birth weight spectrum, 12 studies considered infant macrosomia (defined as birth weight > 4,000 or > 4,500 g). Recognizing the variability in definitions of macrosomia and GWG categories, the committee found that the studies showed a consistent trend for increased risk of macrosomia with increasing GWG. Relative risks were 2–3 for macrosomia in the highest compared to the lowest GWG category.

These results consistently indicate that the relationship of GWG to birth weight applies across the full range of weights and is not limited to the low or high end of the distribution. However, because birth weight is a combination of fetal growth and duration of gestation, studies that separate these two components are more informative.

AHRQ Studies on the Association Between GWG and Weight for Gestational Age

The AHRQ review (Viswanathan et al., 2008) identified 15 studies of SGA that did not stratify by prepregnancy BMI; the studies showed a consistent pattern of diminishing risk of SGA with increasing GWG. It is difficult to provide quantitative estimates of the magnitude of this effect given variable study methods and results, but as for LBW, relative risks were on the order of 2–3 across extreme GWG categories. The six studies that stratified by prepregnancy BMI similarly found that lower GWG was associated with increased risk of SGA births. While methods and results were again variable, the studies did not strongly suggest that prepregnancy BMI modified the relationship between lower GWG and SGA, in contrast to the interpretation in the IOM (1990) report.

In the 10 studies in which GWG and LGA were considered, there was reasonably consistent support for a positive association. For each 1 kg increment in GWG, the relative risk of LGA increased by approximately a factor of 1.1, and comparing the highest to lowest categories of GWG yielded relative risks on the order of 2. The studies that stratified by prepregnancy BMI did not show notable differences in the GWG-LGA association across BMI categories, with only a modest tendency towards a stronger association between higher GWG and LGA among women with lower prepregnancy BMI.

Other Studies on the Association Between GWG and Fetal Growth

Subsequent to the AHRQ review (Viswanathan et al., 2008), four additional studies addressing GWG and birth weight had been published. First, Lof et al. (2008), whose focus was on the role of physical activity in relation to GWG and pregnancy outcome, noted that GWG during weeks 12–33 (unadjusted for prepregnancy BMI) was modestly correlated with increased birth weight (r = 0.13; p = 0.05) and more strongly correlated with birth weight than GWG during either weeks 12–25 or 25–33 alone. Second, Segal et al. (2008) found similar results in a study of obesity and family history of diabetes in relation to pregnancy outcome; controlling for prepregnancy BMI, they reported an adjusted correlation coefficient of 0.19 (p = 0.09) between weight gain before the oral glucose tolerance test and birth weight.

Third, utilizing data from the Danish National Birth Cohort, Nohr et al. (2008) conducted the most informative and detailed analysis to date on the independent effects of prepregnancy BMI and GWG. Analyzing data from over 60,000 births, the authors evaluated the relationship between GWG and both SGA and LGA, as well as the interaction between prepregnancy BMI and GWG in relation to birth weight. They reported statistically significant but generally modest indications of an interaction between prepregnancy BMI and GWG, with the exception of a stronger association of low GWG with SGA among underweight women.

Subsequent analyses of this data (information contributed to the committee in consultation with Nohr) revealed that the relative risk of SGA associated with lower (< 10 kg) versus medium (10–15 kg) GWG among underweight women was 2.1, while it was 1.7 for normal weight women, 1.6 for overweight women, and 1.3 for obese women. The increased risk of LGA associated with very high GWG (≥ 20 kg) vs. medium GWG (10–15 kg) was 3.7 for underweight women, 2.6 for normal weight women, 2.0 for overweight women, and 1.8 for obese women, again suggesting that the effect of lower GWG on risk of SGA is dampened with increasing prepregnancy BMI. This large, carefully done study is important not only because it quantifies the magnitude of effect of GWG on birth weight, but also because it is consistent with the large body of previous evidence demonstrating an overall shift of fewer SGA and more LGA births (and higher mean birth weight) with increasing GWG (see Appendix G, Part I).

Fourth, utilizing data from the Pregnancy Risk Assessment Monitoring System (PRAMS), Dietz et al. (2009) estimated associations between GWG and delivery of an SGA infant using three definitions of SGA: > two standard deviations below the mean birth weight for gestational age, a customized measure, < 10th percentile of expected birth weight for gestational age, and < 10th percentile of birth weight for gestational age using a population-based reference (i.e., information derived from about 104,980 singleton term births in 2000–2005 from 29 states participating in PRAMS). The magnitudes of association between GWG and SGA are striking, with more than a 10-fold gradient in risk from lowest to highest weight gain categories for underweight women, and a 3- to 4-fold gradient in risk for women in the other BMI categories (see Table 6-2). Risk of LGA births or births > 4,500 g yielded clear and similar findings; with increasing weight gain, there was a markedly increased risk of LGA births, present among all BMI groups, but most pronounced on a relative scale among the women with the lowest BMI.

TABLE 6-2

Adjusted Odds Ratios for Association of Total GWG with SGA Stratified by Prepregnancy BMI.

Summary of the Evidence on an Association Between GWG and Fetal Growth

In summary, the evidence that GWG is related to birth weight for gestational age based on observational studies is quite strong and the magnitude of that association is large, with relative risks of SGA with low GWG on the order of 2–3. It appears that the entire birth weight distribution is shifted upward with increased GWG, reducing the risk of SGA and increasing the risk of LGA as the mean birth weight rises. The evidence that this pattern is enhanced among women with low prepregnancy BMI is moderately strong as well.

It is not yet clear, however, whether the associations between GWG and birth weight for gestational age is impacted by factors other than prepregnancy BMI. The IOM (1990) report suggested consideration of a different relationship between GWG and fetal growth among young mothers, but studies conducted since then have failed to provide any additional support for the differential effects by maternal age group. Research on the potentially differential effects of GWG on fetal growth according to ethnicity, smoking status, or other maternal attributes has been sparse, and the few studies summarized in the AHRQ review inconsistent. In addition to prepregnancy BMI, the only other factor that appears to impact the association between GWG and birth weight for gestational age is time during pregnancy that GWG occurs, with modest support for a stronger effect of GWG that occurs during the first or second trimester than during the third trimester GWG (Viswanathan et al., 2008).

Biological Plausibility

Although the pathogenesis of spontaneous preterm delivery has not been clearly elucidated, researchers have postulated at least five possible primary pathogenic mechanisms (IOM, 2007):

1. Activation of the maternal or fetal hypothalamic-pituitary-adrenal (HPA) axis.
2. Amniochorionic-decidual or systemic inflammation.
3. Uteroplacental thrombosis and intrauterine vascular lesions.
4. Pathologic distention of the myometrium.
5. Cervical insufficiency.

The committee found no studies that directly link GWG to activation of the maternal or fetal HPA axis. However, several animal studies have linked periconceptional under nutrition to accelerated maturation of fetal HPA axis resulting in preterm delivery (Bloomfield et al., 2003, 2004; Kumarasamy et al., 2005).

Again, the committee also found no studies directly linking GWG to amniochorionic-decidual or systemic inflammation. However, it is plausible that maternal undernutrition may increase the risk of preterm delivery by suppressing immune functions or increasing oxidative stress. Macro- or micronutrient deficiencies are known to adversely affect maternal immune functions. For example, iron-deficiency anemia can alter the proliferation of T- and B-cells, reduce the killing activity of phagocytes and neutrophils, and lower bactericidal and natural killer cell activity, thereby increasing maternal susceptibility to infections (Allen, 2001). Furthermore, protein and/or micronutrient deficiencies may impair cellular antioxidant capacities because proteins provide the amino acids needed for synthesis of antioxidant defense enzymes, such as glutathione and albumin (reactive oxygen species scavengers); and many micronutrients themselves are antioxidants. Increases in reactive oxygen species, such as oxidized low-density lipoprotein and F2-isoprostanes (lipid peroxidation products), may contribute to cellular toxicity, inflammation, vasoconstriction, platelet aggregation, vascular apoptosis, and endothelial cell dysfunction (Luo et al., 2006), which may also activate the pathway to preterm delivery involving uteroplacental thrombosis and intrauterine vascular lesions.

Summary of the Evidence on an Association Between GWG and Preterm Birth

In summary, there is strong evidence for a U-shaped association between lower GWG and preterm birth among normal weight and under-weight women, and moderate evidence for an association of higher GWG and preterm birth. The magnitude of the association is fairly strong, with relative risks on the order of two, but difficult to summarize because of variability in the definitions of higher and lower rates of weight gain. There is no empirical basis for suggesting modifiers of this relationship other than prepregnancy BMI, for which the data are clear in showing that associations of low GWG with preterm birth are stronger among underweight women.

The committee was unable to infer a causal relationship between GWG and preterm delivery based on available evidence. Although there are intriguing data linking macro- and/or micronutrient deficiencies to accelerated maturation of fetal HPA axis and altered immune functions and/or increased oxidative stress, suggesting that a direct causal relationship is biologically plausibility, important questions regarding timing, threshold, content, and interactions remain unanswered. These uncertainties about a direct causal relationship between GWG and preterm delivery guided the committee’s approach to decision analysis in Chapter 7, which weighed the trade-offs of GWG with and without taking into account preterm delivery as an outcome.

Breastfeeding Outcomes

Breastfeeding is an important outcome to study not only because it may be associated with reduced offspring obesity, and therefore may serve as an intermediate marker such as infant weight gain, but also because it predicts other health outcomes such as reduced otitis media, gastrointestinal illness, and better cognition. Although observational studies (see Chapter 5) have documented a relationship between excessive weight gain during pregnancy and poor breastfeeding outcomes, the committee identified no studies that addressed the relationship between GWG and lactation-related offspring outcomes.

Summary of the Evidence on an Association Between GWG and Childhood Obesity

In summary, the evidence to date is suggestive but not conclusive that GWG outside the ranges recommended by IOM (1990) is associated with higher offspring BMI. Evidence that the association is effected by maternal BMI is scant, and there is no evidence on whether the timing of weight gain during pregnancy impacts offspring BMI. Most of the studies rely on BMI as the only outcome, in contrast to direct measures of adiposity and cardio-metabolic status, which would strengthen the evidence base. Only one study to date has reported on blood pressure as an outcome (Oken et al., 2007), although another recent report suggests that higher weight gain is associated with an increase in left ventricular mass from birth to 6 months (Geelhoed et al., 2008).

Nevertheless, as discussed in the chapter opening, even strong observational studies that have valid exposure and outcome measures, large sample sizes, and appropriate control for confounding cannot fully address the question of causality. Randomized controlled trials that are designed to modify GWG and include follow-up of the children would provide the most compelling evidence for or against intensive clinical or public health efforts to curb excessive weight gain.

Neurodevelopment

Alterations in fuel metabolism during pregnancy resulting from intended or unintended weight loss, fasting, or poorly controlled diabetes can cause ketonemia and/or ketonuria, which in turn can have consequences for the neurocognitive development of the infant (see Chapter 3). The committee reviewed the evidence for long-term neurodevelopmental consequences of ketonemia/ketonuria in pregnancy (see Appendix G). As a result of the association between lower GWG and SGA (see discussion in the Fetal Growth section of this chapter), one indirect way to evaluate the impact of GWG on neurodevelopment is by assessing associations of term and preterm SGA with neurodevelopmental outcomes.

Long-term neurodevelopment in term SGA The committee was able to identify only observational prospective studies and review articles evaluating long-term neurodevelopmental outcomes in term SGA infants. With the exception of one study conducted in China, all were conducted in industrialized nations including the United States (Goldenberg et al., 1996; Nelson et al., 1997). Of 18 studies identified, 6 examined neurodevelopmental outcomes during infancy/childhood (Watt and Strongman, 1985; Nelson et al., 1997; Sommerfelt et al., 2000; Hollo et al., 2002; Geva et al., 2006; Wiles et al., 2006), 9 during adolescence (Westwood et al., 1983; Paz et al., 1995, 2001; Pryor et al., 1995; Goldenberg et al., 1996; Strauss, 2000; O’Keefe et al., 2003; Indredavik et al., 2004; Peng et al., 2005; Kulseng et al., 2006) and 2 during adulthood (Viggedal et al., 2004; Wiles et al., 2005). Results have been mixed.

A study from a Finnish cohort found that SGA children performed worse at school than gestational age-matched controls at 10 years of age (Hollo et al., 2002). Another cohort of children in Norway found a slightly lower mean intelligence quotient (IQ) at 5 years of age associated with SGA; however, parental factors were more strongly related with IQ than SGA (Sommerfelt et al., 2000). Nelson et al. (1997) found that SGA was not associated with either the Bayley Mental Development Index (MDI) and Fagan Test of Infant Intelligence score at 1 year of age; in contrast, they found that SGA was associated with a lower Bayley Psychomotor Development Index (PDI) among black males but not among females or white male and female infants. Watt and Strongman (1985) documented that SGA was inversely associated with MDI developmental scores at 4 months, whereas Goldenberg et al. (1996) found an inverse relationship between SGA and IQ at 5.5 years of age. Wiles et al. (2006) did not find a relationship between low birth weight and behavioral problems at 6.8 years of age.

Effect size analysis was assessed for cognitive outcome measures. Of the 18 studies reviewed, 12 reported cognitive scores by SGA status. Of these, 2 reported lower Bayley scores (Watt and Strongman, 1985; Nelson et al., 1997) and 10 reported lower IQ measures (Westwood et al., 1983; Paz et al., 1995, 2001; Pryor et al., 1995; Goldenberg et al., 1996; Sommerfelt et al., 2000; Hollo et al., 2002; Viggedal et al., 2004; Peng et al., 2005; Kulseng et al., 2006) associated with term-SGA status, although these differences were not always statistically significant. Among infants, the Bayley score difference associated with SGA ranged from 4–7 points (Watt and Strongman, 1985; Nelson et al., 1997). Among children (Goldenberg et al., 1996; Sommerfelt et al., 2000; Hollo et al., 2002), IQ differentials were 4–5 points. Among adolescents (Westwood et al., 1983; Paz et al., 1995, 2001; Pryor et al., 1995; Peng et al., 2005; Kulseng et al., 2006) the corresponding range was 2–12 points. The upper range limit was derived from a study conducted in China that did not control for socioeconomic and other potential confounding factors (Peng et al., 2005). The only study among adults that reported IQ documented a relatively large 19-point IQ differential (based on scores’ median instead of average) associated with term-SGA infants. However, it also failed to account for confounding factors (Viggedal et al., 2004). Overall, these 18 studies consistently reported small cognitive differentials associated with being born at term and SGA. The meaning of these small differentials is unclear, as in all studies average scores among individuals born SGA fell within the normal IQ range.

Associations between SGA and long-term neurodevelopmental outcomes among term newborns were inconsistent, especially among adolescents. Among studies that supported this association, major methodological shortcomings (which included substantial attrition, lack of standard definitions of SGA across studies, not properly accounting for key confounders [such as socioeconomic status and parental cognitive functioning, as well as asphyxia at birth], and lack of testing for effect modification by environmental factors) limit their interpretation. As a whole, the studies were not designed to identify the influence of SGA, independent of socioeconomic factors, on lower IQs. The committee’s evaluation of the evidence concurs with previous reviews (Grantham-McGregor, 1995; Goldenberg et al., 1998) that SGA is associated with minimal neurologic dysfunction (e.g., poor school performance) and is not associated with major handicaps, such as cerebral palsy, unless accompanied by asphyxia at birth.

Long-term neurodevelopment in preterm SGA In preterm SGA infants, the majority of longitudinal studies reviewed by the committee focused on extremely premature (Feldman and Eidelman, 2006; Kono et al., 2007; Paavonen et al., 2007; Leonard et al., 2008) or very low birth weight (VLBW) (Litt et al., 1995; Hack, 1998; Brandt et al., 2003; Kilbride et al., 2004; Litt et al., 2005; Feldman and Eidelman, 2006; Hille et al., 2007; Paavonen et al., 2007; Strang-Karlsson et al., 2008a, 2008b) infants. Among 14 studies in children, 11 found that SGA was associated with cognitive and/or neurodevelopment impairments, although this relationship may be modified by degree of postnatal catch-up growth and maternal-child interactions (Casey et al., 2006; Feldman and Eidelman, 2006). In general, the effect size was proportional to the severity of prematurity (Calame et al., 1983; Feldman and Eidelman, 2006; Kono et al., 2007). The two studies conducted among adolescents found an association of VLBW with IQ (Hille et al., 2007) and breathing-related sleep disorders (Paavonen et al., 2007). Among adults, VLBW was associated with emotional instability (Strang-Karlsson et al., 2008b) and SGA with lower head circumference among individuals who did not fully catch up in their head circumference growth during their first 12 months of life.

Effect size was again assessed for cognitive measures. Of 19 studies reviewed, 13 reported cognitive scores by SGA status; of these, 1 reported a lower Bayley score (Feldman and Eidelman, 2006) and 12 reported lower IQ measures (Escalona, 1982; Calame et al., 1983; Silva et al., 1984; Holwerda-Kuipers, 1987; Litt et al., 1995; McCarton et al., 1996; Hutton et al., 1997; Kilbride et al., 2004; Litt et al., 2005; Casey et al., 2006; Hille et al., 2007; Kono et al., 2007) associated with preterm SGA status, although these differences were not always statistically significant. Among 2-year-old children, one study found an 8-point difference in the Bayley Mental Development Index score (Feldman and Eidelman, 2006). In contrast, a study conducted among 3.5-year-old children found no differences in IQ scores associated with preterm SGA (Escalona, 1982). Among the rest of studies with children (Calame et al., 1983; Silva et al., 1984; Holwerda-Kuipers, 1987; Litt et al., 1995; McCarton et al., 1996; Hutton et al., 1997; Kilbride et al., 2004; Litt et al., 2005; Casey et al., 2006; Kono et al., 2007), IQ differentials were 2–11 points. The only study among adults that reported IQ, documented a 2-point differential associated with VLBW. Overall, the cognitive differentials appear to be relatively stronger among individuals born SGA preterm (mean ± std. dev: 6.5 ± 3.8 IQ points, n = 11 studies) than among those born SGA term (5.3 ± 3.0, n = 9 studies IQ points). However, as with term SGA, the meaning of these still relatively small differentials is unclear because in the vast majority of studies the average scores for individuals born preterm SGA fell within the normal IQ range.

The overwhelming majority of studies reviewed support an association between preterm SGA and lower neurodevelopment in the longer term. Consistent with the studies on term SGA, many of the studies on preterm SGA did not properly control for key perinatal (e.g., asphyxia), socio-economic, parental, and home environment confounders (e.g., maternal-child interactions). In addition, although some studies included term births as reference groups (Calame et al., 1983; Silva et al., 1984; Holwerda-Kuipers, 1987; Litt et al., 1995, 2005; Hack et al., 1998; Brandt et al., 2003; Kilbride et al., 2004; Paavonen et al., 2007; Leonard et al., 2008; Strang-Karlsson et al., 2008a, 2008b), others used preterm subgroups as comparison groups (McCarton et al., 1996; Hutton et al., 1997; Casey et al., 2006; Kono et al., 2007). Because of these study design limitations, the effect size or the proportion of the variance in neurodevelopmental outcomes that can be attributed to being born premature per se or to the combination of prematurity and SGA still needs to be determined.

In summary, as was the case with infant mortality, one must link GWG to being born preterm or small- or large-for-gestational age and, from there, to neurodevelopmental outcomes. This sequence is biologically plausible and it is possible that it is causal, but the evidence to establish causality is not available.

Apgar score The Apgar score (see Glossary in Appendix A) assessments are usually conducted 1 and 5 minutes after birth, and scores can range from 0 to 10. However, Apgar scores in term infants, even at 5 minutes, have important limitations, as they are not adequate predictors of longer term morbidity and mortality and do not correlate well with neurological outcomes (ACOG, 2006) although very low scores (0–3) associated with low birth weight do predict neonatal mortality. The AHRQ review (Viswanthan et al., 2008) identified five studies examining the influence of GWG on a newborn’s Apgar score (Stevens-Simon and McAnarney, 1992; Nixon et al., 1998; Cedergren et al., 2006, Stotland et al., 2006; Wataba et al., 2006). Taken together, these studies provide only modest evidence that excessive GWG is associated with low Apgar score, and one study suggested that low GWG in nulliparous women also predicts low Apgar score.

Childhood cognition No published studies directly examine the link between GWG and neurocognitive development in infants and children. However, as discussed in Chapter 3, weight loss or failure to gain during pregnancy due to dietary caloric insufficiency may possibly induce maternal hormonal and metabolic responses, which may, in turn, have subsequent consequences for the intellectual development of the child. Because of the obligatory weight gain in maternal tissues (uterus, breast, blood) and the fetal-placental unit, a weight gain less than ~7.5–8.5 kg would likely result in mobilization of maternal adipose tissue and possibly lean body mass. Although the gestational metabolic milieu or offspring outcomes of pregnant women who experience weight loss have not yet been addressed in the scientific literature, there have been several studies on associations between ketonemia or ketonuria, which can occur among pregnant women subjected to short-term fasting (see Chapter 3), and cognition in offspring. Some, but not all, of these studies have found an association between biomarkers of maternal metabolic fuel alterations and child intellectual development (Stehbens et al., 1977; Rizzo et al., 1991; Silverman et al., 1991). In contrast, Persson and Gentz (1984) and Naeye and Chez (1981) did not find any association of maternal acetonuria, weight loss, or low GWG with either psychomotor development or IQ in children (see Chapter 3).

In summary, although no studies specifically address the impact of very low GWG or weight loss on child intellectual development, some evidence suggests that biomarkers of short-term negative energy balance during pregnancy may be related to the child’s intellectual development. These associations may be limited to women with diabetes during pregnancy.

Allergy/Asthma

Inasmuch as GWG has been associated with risk of preterm birth, it is plausible that GWG may also be a risk factor for childhood asthma since prematurity itself is a risk factor for childhood asthma—often as a result of suboptimal lung function and resulting neonatal respiratory morbidity (Dombkowski et al., 2008). For example, in a case-control study of 262 African American 4- to 9-year-old children receiving care at a hospital-based clinic, Oliveti et al. (1996) used maternal GWG (as determined from the mother’s self-report of prepregnancy weight and maternal weight measured at the time of delivery) and child medical records to examine pre- and perinatal risk factors for asthma (as defined either by physician diagnosis or wheezing or coughing that required asthma medication).

Multivariate logistic regression analyses showed that odds of prevalent asthma were 3.42 (95% CI: 1.72–6.79) times higher among women who gained less (versus more) than 20 total pounds during pregnancy. However, the authors neither adjusted for prepregnancy BMI nor examined the BMI-GWG interaction. Among children with asthma, 24.6 percent were born preterm compared to 13.7 percent of controls.

Another way that GWG could lead to the propensity for asthma in offspring is through alteration of the developing fetal immune system. For example, Willwerth et al. (2006) found that both inadequate and excessive GWG were associated with increased cord blood mononuclear cell proliferative responses to stimulation (OR = 2.3 and 2.6, respectively), compared to controls with adequate GWG. In that study, maternal smoking (OR = 18) was the major determinant of the response.

Cancer

Whether associations exist between GWG, birth weight, and risk for childhood cancers is not clear; however, there are a few studies that have examined the possibility.

Childhood leukemia Two lines of evidence link GWG to cancer: First, a recent meta-analysis (Hjalgrim et al., 2003) estimated that the odds for acute lymphoblastic childhood leukemia (ALL) were higher (OR = 1.26, 95% CI: 1.17–1.37) for infants with birth weight over 4,000 g compared to those under 4,000 g. Although not statistically significant, results were of similar magnitude for acute myelogenous leukemia (AML). McLaughlin et al. (2006) examined the association between pregnancy outcomes and leukemia cases registered in the state of New York and diagnosed before 10 years of age. The authors obtained information on prepregnancy weight (BMI not generally available) and GWG from birth certificates. Using multivariate regression analyses, the researchers found that total GWG greater than 14 kg conferred an increased risk for ALL (OR = 1.31, 95% CI: 1.07, 1.60). Maternal prepregnancy weight did not affect the impact of GWG on ALL; nor was there any association between GWG and AML. The authors speculated that higher GWG could result in higher fetal exposure to insulin-like growth factor I (IGF-I), which in turn may increase the risk of childhood ALL. Based on these results and because of the established relationship between higher GWG and macrosomia (see discussion in Fetal Growth section in this chapter), it is plausible that GWG may be related to childhood leukemia.

Breast cancer Almost two decades ago, Trichopoulos (1990) hypothesized that breast cancer originates from alterations in the prenatal endocrine milieu, in particular higher estrogen levels. Although longitudinal studies needed for definitively testing this hypothesis have yet to be conducted, observational studies showing direct associations between birth weight and breast cancer provide some support for an association (Michels et al., 1996; Vatten et al., 2002; Ahlgren et al., 2007). Therefore, the committee deemed it of interest to examine determinants of hormone levels in the maternal-placental-fetal unit. In a sample of 270 white women from Boston, Lagiou et al. (2006) examined the association between GWG and maternal sex hormones at 16 and 27 weeks of gestation. After adjusting for prepregnancy BMI and other covariates the authors did not detect any associations between GWG and maternal estradiol, estriol, or prolactin levels, providing some support to the estrogen hypothesis; however, higher GWG was associated with lower levels of maternal progesterone and of sex hormone-binding globulin (−0.7 percent [95% CI: −1.5, 0.0] at 16 weeks and −1.2 percent [95% CI: −2.0, −0.4] at 27 weeks, respectively, for every 1-kg increment in GWG.

In addition, one study directly addressed the association of GWG with incident breast cancer. Analyzing data from the Finnish Cancer Registry, Kinnunen et al. (2004) found that offspring of mothers in the upper tertile of GWG (> 15 kg) had a 1.62-fold higher breast cancer risk than mothers who gained within the recommended range (11–15 kg) after adjusting for parity; mother’s age at menarche, at first birth, and at index pregnancy; and prepregnancy BMI. Together these findings provide some support for the hypothesis that excessive weight gain in pregnancy could lead to elevated breast cancer risk in the offspring.

Attention deficit hyperactivity disorder Because the human brain develops rapidly during both gestation and the early postnatal period, it is possible that maternal body fat reserves and GWG can influence fetal central nervous pathways in a manner that ultimately results in long-term behavioral disorders such as attention deficit hyperactivity disorder (ADHD). Only one study was able to identify associations between either early pregnancy BMI or GWG and ADHD (Rodriguez et al., 2008). Within three cohorts of 7- to 12-year-old Scandinavian children, teachers used standard questionnaires to rate the children’s inattention and hyperactivity symptoms; about 8.5 percent of children were classified as having ADHD symptoms. A large majority of the mothers (86.4 percent) had a normal prepregnancy BMI and adequate GWG (mean gestational age = 39.6 ± 1.6 weeks, mean birth weight = 3.6 ± 0.5 kg). Multivariate linear regression analyses showed that among women with a high prepregnancy BMI, GWG (measured weekly and in 100-g increments) was associated with increased odds of child ADHD (OR = 1.24, 95% CI: 1.07–1.44). Among lean women, those who experienced weight loss during pregnancy also had higher odds of having a child that would later develop ADHD compared to their counterparts who did not lose weight during gestation (OR: 1.52, 95% CI: 1.07–2.15). The mechanisms for these effects are unknown, although the authors speculated on the possibility of neurotoxin transfer from maternal adipose tissue to the developing fetal brain.