The Association Between Maternal Glucose Concentration and Child BMI at Age 3 Years

Subjects were recruited from the third cohort of the Pregnancy Infection and Nutrition study (PIN), January 1, 2001 through June 30, 2005, and followed through 3 years postpartum by the PIN Postpartum (3 and 12 months postpartum) and PIN Kids (3 years postpartum) studies. The recruitment protocols for the studies were documented previously (8). Briefly, pregnant women who were at least 16 years of age at conception, English speaking, before 20 weeks' gestation at recruitment, and receiving prenatal care from public and private clinics at the University of North Carolina (UNC) Hospitals were eligible for participation. The PIN study protocols were approved by the Institutional Review Boards of the School of Medicine at UNC Chapel Hill.

A total of 1,169 women in PIN delivered a live singleton infant and were eligible for recruitment into the postpartum study. There were 689 and 550 mother-child pairs who participated at 3 and 12 months postpartum, respectively. The most common reasons for nonparticipation included refusal/request to leave study (n = 202), unreachable/moved out of study catchment area (n = 226), and ineligibility (n = 338) because of medical constraints, timing issues, or pregnancy. Recruitment for PIN Kids began in 2004 and added an assessment of the index infant at 3 years. There were 409 mother-child pairs who completed the 3-year visit. Three children were not eligible for inclusion because a physician diagnosed growth-related illness. Anthropometric measurements were missing for 81 children, mostly because other data were collected by phone interview rather than home visit (n = 58) or the child was unavailable, such as napping, during the home interview. We further excluded mothers with pre-existing diabetes (n = 20), GDM (n = 9), or missing glucose concentration information (n = 3). Children born before 37 weeks' gestation (n = 30) were also excluded. The remaining 263 mother-child pairs were included in the current analyses.

Distributions of selected baseline characteristics between eligible mother-child pairs who participated in PIN Kids (n = 406), PIN Postpartum (at 3 months postpartum) pairs not included in PIN Kids (n = 283), and eligible PIN pairs not participating in PIN Postpartum (n = 480) were examined. In comparison with mothers in PIN Kids, those not participating in both PIN and PIN Postpartum were significantly younger and had infants with lower mean birth weight and gestational age. They were more likely to be obese, black, unmarried, less educated, and from low income households. Additionally, a higher percentage of PIN Postpartum mothers not participating in PIN Kids smoked during pregnancy. All other comparisons of characteristics were not significant. There were no significant differences in the baseline characteristics of the PIN Kids samples included (n = 263) in this analysis and those excluded (n = 143) with the exception that excluded women had a slightly higher prepregnancy BMI (mean SD). BMI was 26.3 (7.8) kg/m² and 24.3 (5.9) kg/m² for excluded and included women, respectively (P = 0.004).

At the 3-year home visit, children’s heights and weights were measured by trained PIN staff using stadiometers and scales, respectively, according to National Health and Nutrition Examination Surveys protocols (9). The mean (SD) age of the children at the home visit was 3.04 (0.15) years. Child BMI at 3 years was calculated from these measurements (in kg/m²) and converted to age- and sex-specific BMI z scores (continuous) and percentiles (categories) using the 2000 Centers for Disease Control and Prevention/National Center for Health Statistics growth charts (10). Overweight/obesity was defined as BMI-for-age and sex ≥85th percentile (reference <85th percentile).

Maternal plasma glucose concentrations were derived from universal screens administered during the second trimester (mean gestational age ∼27 weeks) on all women without pre-existing diabetes. The universal screen involved administration of a random 1-h 50-g glucose challenge test (GCT) to the women. Women with abnormal values on the GCT were administered a 3-h oral glucose tolerance test to confirm GDM using Carpenter and Coustan criteria (11). For this analysis, maternal plasma glucose concentrations from the GCT were categorized using the following cut points: <100, 100–<130, and ≥130 mg/dL. These cut points are comparable with those used in a previous study (12), and glucose concentrations ≥130 mg/dL are suggestive of hyperglycemia (11).

Maternal prepregnancy BMI (in kg/m²) was calculated using self-reported prepregnancy weight and measured height. For quality assurance, weight measurements taken at the first prenatal clinic visit (within 15 weeks' gestation) were compared with the self-reported prepregnancy weights to identify biologically implausible weight gains. Women with implausible values (n = 7) had their prepregnancy weights imputed following previously published methods (8). Maternal prenatal smoking (months 1–6 of pregnancy), household income, education, marital status, age at conception, race/ethnicity, pre-existing diabetes, and parity were collected from prenatal interviews and categorized as shown in Table 1. Household income was expressed as percentage of the poverty line and calculated using the 2001 U.S. Department of Health and Human Services Federal Poverty Guidelines; a percentage ≤185% is the cutoff for the Special Supplemental Nutrition Program for Women, Infants, and Children. Household income at the 3-month postpartum interview was used for missing information from the prenatal period (n = 7). Infant sex and birth weight were recorded at delivery and abstracted from the medical records. Gestational age was determined from ultrasound measurements conducted before 22 weeks gestation (up to 21 weeks, 6 days). If no ultrasound was performed or if it was not performed before the start of the 22nd week then the date of the last menstrual period was used (n = 5). Birth weight z scores specific for infant sex and gestational age were calculated using U.S. national reference data (13).

Statistical analyses were performed using STATA 11 (College Station, TX). Potential effect measure modifiers, confounders, and mediators of interest were identified a priori from a literature review and causal diagrams (14). The interaction of maternal glucose concentration and prepregnancy BMI was tested in the full models using interaction terms and Wald tests with an a priori significance P value of <0.15. Student t tests, Fisher exact tests, and analyses of variance were used to analyze distributions of baseline characteristics. Multivariable linear regression models were used to examine associations of continuous BMI z scores with maternal glucose concentration. Modified Poisson regression (Poisson regression with a robust error variance) estimated risk ratios of child overweight/obesity (BMI ≥85th percentile). This method has been validated for directly estimating relative risks for dichotomous, common outcomes in prospective studies (15). All regression analyses were adjusted for clustering at the individual level (16) since there were five women with more than one child included in the analyses. The analyses were repeated excluding the second child of these women, and there were no appreciable differences in the results.

The mean (SD) maternal glucose concentration was 103.8 (23.7) mg/dL. The mean (SD) prepregnancy BMI was 24.3 (5.9) kg/m². Approximately 30% of the women were overweight (17.6%) or obese (12.2%). The majority of women were 25–34 years at conception, nonblack, married, achieved a high school degree or higher, upper income (>350% of the 2001 Poverty Guidelines), and nonsmokers during pregnancy. Approximately half of them were nulliparous. The mean (SD) birth weight and gestational age of the children were 3415.8 (413.4) g and 39.2 (1.2) weeks, respectively. At 3 years, the mean (SD) child BMI z score was 0.29 (0.99), 20.9% were overweight/obese (BMI ≥85th percentile), and 5.3% were obese (BMI ≥95th percentile). Table 1 shows the distributions of selected characteristics for the sample (n = 263) according to mean child BMI z score. Mean child BMI z score significantly differed across categories of maternal glucose concentration and prepregnancy BMI, as well as maternal prenatal marital status, education, and smoking.

Table 2 displays the results from linear regression analyses for the association of maternal glucose concentration categories and child BMI z score at 3 years. Each 1 mg/dL increase in maternal glucose concentration was associated with a slight increase in BMI z score at 3 years (estimated difference in z score of 0.005, 95% CI: −0.001, 0.010, P = 0.08 in the adjusted model, data not shown in table). When compared with glucose concentration <100 mg/dL, a concentration ≥130 mg/dL was associated with significantly higher child BMI z score at 3 years in the adjusted model (estimated difference in z score of 0.39, 95% CI: 0.03–0.75). We also examined the association of maternal glucose concentration categories and risk of child overweight/obesity at 3 years (Table 3). In comparison with glucose concentration <100 mg/dL, concentration ≥130 mg/dL was associated with an approximate twofold greater risk of child overweight/obesity (adjusted risk ratio 2.34, 95% CI: 1.25–4.38). Additional adjustment of models for infant birth weight z score did not substantially alter any of the effect estimates. We also did not find an interaction between maternal glucose concentration and prepregnancy BMI; however, because of the small sample size we may have lacked adequate power to detect one if it existed.

With the use of data from mother-child pairs participating in a recent, prospective, longitudinal cohort study, we found a significant increased risk of child overweight/obesity at 3 years associated with maternal glucose concentration ≥130 mg/dL. These results suggest that fetal exposure to high maternal glucose concentration in the absence of pre-existing diabetes or GDM may contribute to the development of overweight/obesity in the offspring, independent of maternal prepregnancy BMI.

Nearly 60 years ago, Pedersen (17) hypothesized that fetal exposure to maternal hyperglycemia via diabetes resulted in fetal hyperinsulinemia and greater adiposity at birth. Since then both animal and human studies provided evidence to support Pedersen’s observations and increased our understanding of the pathophysiology of GDM (18). GDM is associated with several adverse offspring outcomes including macrosomia; large-for-gestational age (LGA); respiratory distress (1); insulin resistance; and greater risk of obesity, type 2 diabetes, and metabolic syndrome later in life (1). There is adequate literature to suggest that the associations for early infant outcomes (macrosomia [19], LGA [12], and adiposity [3]) extend to hyperglycemia during pregnancy in the absence of pre-existing diabetes or GDM. However, there are fewer studies examining associations with later childhood outcomes, specifically obesity.

Consistent with the current study, two studies found positive associations for maternal glucose concentration and offspring anthropometric outcomes. Among nondiabetic Pima Indians, Pettitt et al. (4) showed a linear association between maternal glucose concentration (2-h, 75-g oral glucose tolerance test [OGTT]) and offspring relative weights at ages 5–24 years. There was also a higher prevalence of obesity and higher rates of abnormal glucose tolerance in offspring of mothers who had abnormal glucose tolerance during pregnancy. In a multiethnic U.S. population, Hillier et al. (5) reported significant increased odds of child BMI >85th percentile (adjusted odds ratio [aOR]: 1.22, 95% CI: 1.03–1.45) and BMI >95th percentile (aOR: 1.28, 95% CI: 1.02–1.60) at 5–7 years associated with maternal glucose concentration (1-h, 50-g GCT) of 122–140 mg/dL compared with 43–94 mg/dL.

In contrast, using data from the Belfast, U.K. center of the HAPO study, which was designed to clarify the risks of adverse outcomes associated with various degrees of maternal glucose intolerance less severe than that in overt diabetes mellitus (3), Pettitt et al. (6) found no association between maternal glucose concentration (assessed using fasting glucose and 1-h and 2-h OGTT) and child BMI ≥85th and ≥95th percentiles at 2 years. The only exception was a significant increase in prevalence of child BMI ≥85th percentile across strata of maternal 1-h glucose concentration. Similarly, in a nondiabetic homogeneous Caucasian population in Exeter, U.K. (20), despite finding significant correlations between fasting plasma glucose (measured at 28 weeks' gestation) and offspring weight, length, and BMI at birth, none of the estimates remained significant when measurements were repeated at ages 12 weeks, 1 year, and 2 years.

It has been suggested that the effects of maternal glucose concentration are long term and differ depending on the age of the offspring (6). For example, when comparing offspring of women with and without GDM, there are significant differences in weight at birth, which are not apparent at 1.5 years but then recur later in childhood at 7.7 years (21). In this study, children were ∼3 years, which is younger than the ages of the populations used in studies that reported positive associations (4,5) but older than those used in studies reporting null findings (6,20). Our results add to the current evidence suggesting that the effects of fetal glucose exposure on the development of child overweight/obesity are apparent at later ages. However, there is a lack of consistency across studies in methodology concerning maternal glucose concentration measurements, study populations, as well as adjustment for potential confounding variables, which makes it difficult to compare results. More research is necessary to understand the influence of maternal glucose concentration on offspring anthropometric development among women without diabetes.

There are several limitations of the current study that must be considered when interpreting the results. A main limitation is the loss to follow-up between birth and 3 years that resulted in a disproportionate loss of women from high risk groups and likely weakened the observed associations. Additionally, the loss in sample size may have diminished the power necessary to detect some associations and interactions, if they existed. The generalizability of our sample may be limited because women were mostly white, well educated, and upper income, resulting in a lower prevalence of overweight and obesity compared with national estimates for women of reproductive ages (22).

Other limitations are due to our measurements of maternal glucose concentration and child anthropometrics. A standard cut-point of 140 mg/dL defined an abnormal value on the universal screen and resulted in further testing for GDM; however, a cut-point of 130 mg/dL is more sensitive and may be considered a better diagnostic tool for GDM (23). Therefore, some of the women with GCT values ≥130 mg/dL but <140 mg/dL (n = 20) may have been missed as GDM cases and included in our analyses. This type of misclassification would likely strengthen the observed effect estimates, especially if these women remained untreated. Child BMI status is not a direct measure of adiposity but it is correlated (r = 0.75) with percent body fat in children aged 3–8 years (24). Finally, it is likely that the observed associations between maternal glucose concentration and child BMI partly reflect the effects of genetic or environmental factors (such as child diet and physical activity) related to a woman’s glucose concentration and health behaviors in the postnatal period, which we were unable to consider in our etiologic model.

The results from this study suggest that high maternal glucose concentration (≥130 mg/dL) in the absence of a pre-existing diabetes or GDM diagnosis is a modifiable and preventable behavioral factor associated with offspring overweight/obesity at 3 years. Prenatal intervention efforts may consider monitoring women who meet such criteria as a way to decrease fetal exposure to high maternal glucose concentration and possibly prevent excess weight development. Future studies, such as the HAPO study (2), that collect longitudinal data before pregnancy through late childhood, including comprehensive measures of maternal glucose concentration, child adiposity, and potential confounders, are needed.

This study received support from the National Institute of Child Health and Human Development, National Institutes of Health (HD-37584, HD-39373), the National Institute of Diabetes and Digestive and Kidney Diseases (DK-61981, DK-56350), and the Carolina Population Center.

No potential conflicts of interest relevant to this article were reported.

A.L.D. was responsible for the statistical analysis and writing of the article. A.M.S.-R. was a principal investigator of the study and guided analysis and writing of the article. K.C. contributed to the statistical analysis and reviewed and edited the article. A.H.H. was a co-investigator of the study, guided the statistical analysis, and reviewed and edited the article.

The authors would like to thank all of the PIN Study investigators: Kelly Evenson, University of North Carolina at Chapel Hill; Nancy Dole, University of North Carolina at Chapel Hill; David Savitz, Brown University; June Stevens, University of North Carolina at Chapel Hill; and John Thorp, University of North Carolina at Chapel Hill. They obtained funding and designed the study. Additionally, the authors would like to thank Kathryn Carrier, University of North Carolina at Chapel Hill, for managing the PIN study.