February 1998, VOLUME101 /ISSUE 2

Gestational Diabetes and the Risk of Offspring Obesity

  1. Robert C. Whitaker, MD, MPH*,
  2. Margaret S. Pepe, PhD,
  3. Kristy D. Seidel, MS,
  4. Jeffrey A. Wright, MD§,
  5. Robert H. Knopp, MD
  1. From the *Department of Pediatrics, Children's Hospital Medical Center and the University of Cincinnati College of Medicine, Cincinnati, Ohio;
  2. Biostatistics Program, Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington; and the Departments of
  3. §Pediatrics and
  4. Medicine, University of Washington School of Medicine, Seattle, Washington.


Background. Intrauterine exposure to the metabolic alterations of maternal diabetes may increase the risk of later obesity. We determined whether offspring of mothers with diet-treated, gestational diabetes mellitus (GDM) have an increased risk of childhood obesity and examined the relationship between childhood obesity and metabolic markers of GDM.

Methods. At a health maintenance organization in Seattle, WA, we reviewed medical records to obtain the life-time height and weight measurements of 524, 8- to 10-year-old children whose mothers had been screened for GDM. Maternal plasma glucose and triglyceride levels were obtained in midgestation 1 hour after ingestion of 50 g of glucose. Those with glucose screening levels ≥7.77 mmol/L (140 mg/dL) underwent a 3-hour, 100-g, oral glucose tolerance test to determine GDM status. Cord serum insulin levels also were obtained at birth. Obesity was defined as an average body mass index between 5 and 10 years of age at or above the 85th percentile for age and sex.

Results. The prevalence of obesity was 19% in the 58 offspring of mothers with diet-treated GDM and 24% in the 257 offspring of mothers with negative glucose screen values. There also was no difference in mean body mass index (adjusted for age and sex) between these two groups of offspring. Among all 524 offspring, there was no significant increase in the rate of offspring obesity according to the quartile of maternal screening glucose, triglyceride, oral glucose tolerance test, or cord serum insulin level.

Conclusion. Prenatal exposure to the metabolic effects of mild, diet-treated GDM does not increase the risk of childhood obesity.

  • pregnancy in diabetes
  • obesity
  • fetus
  • child
  • body mass index

Children born to mothers with insulin-dependent diabetes mellitus (IDDM) have an increased risk of later obesity.1 These children may be programmed in utero for later obesity by exposure to excess metabolic substrate at a sensitive period in development.6 Gestational diabetes mellitus (GDM), that is, diabetes with onset or first recognition during pregnancy,9 affects 3% to 5% of all pregnancies.10,11 GDM is far more common in pregnant women than is existing IDDM or noninsulin-dependent diabetes mellitus (NIDDM). Previous studies suggest that GDM also increases the risk for later obesity in the offspring.12 Each of these studies, however, contained a mixture of subjects with GDM and either IDDM or NIDDM. No study has ever compared childhood obesity rates in offspring of mothers with and without GDM. Therefore, we tested the hypothesis that offspring from pregnancies affected by GDM have an increased risk of childhood obesity. To suggest a possible mechanism by which the intrauterine environment affects later obesity risk, we also examined the association between childhood obesity and the metabolic markers of GDM measured during gestation.


The study was conducted at Group Health Cooperative of Puget Sound (GHC), a staff-model, health maintenance organization (HMO) in Washington state. In 1985 and 1986, 977 pregnant women at GHC participated in a study of screening tests for GDM.16,17 The offspring of these pregnancies, followed up at 5 to 10 years of age, are the subjects of the current study. The study was approved by the Human Subjects Review Committees both at the University of Washington and at GHC.

GDM Screening Study

At between 24 and 32 weeks' gestation, plasma glucose and triglyceride levels were obtained from each mother 1 hour after a 50-g, oral glucose load. Those with glucose screening values ≥7.77 mmol/L (140 mg/dL) were recalled for a 3-hour, 100-g, oral glucose tolerance test (OGTT). At delivery, an attempt was made to obtain serum insulin levels from the cord blood of all infants.

All 3517 women who enrolled for prenatal care at the two main GHC prenatal clinics between January 1985 and May 1986 were potentially eligible for the GDM screening study (Fig 1). A total of 2019 women without known diabetes consented to participate. Of these, 1477 had a normal glucose screen value, 456 had an abnormal screen value, and 86 were excluded for reasons such as twin gestation or delivery at a non-GHC hospital. A control group of 521 mothers was selected randomly from those 1477 with a normal glucose screen value. The 456 mothers with an abnormal glucose screen value were placed in one of three other groups based on the OGTT result. GDM was diagnosed in 101 mothers, based on any two of the four OGTT values exceeding the criteria published by Carpenter and Coustan18; 264 had a normal OGTT result, and 91 refused to undergo the OGTT. All with GDM were generally prescribed a 7560 to 9240 J (1800- to 2200-kcal) diet low in oligosaccharide, and they were instructed in home glucose monitoring. Women began to receive insulin if they had fasting glucose values >5.9 mmol/L (105 mg/dL) and/or 2-hour postprandial glucose values >6.7 mmol/L (120 mg/dL) on two or more occasions within a 2-week period. Of those mothers with GDM, only 5 were treated with insulin. Two of these five were begun on insulin therapy based on elevated glucose screening values (9.3 mmol/L [168 mg/dL] and 11.0 mmol/L [198 mg/dL]) and did not have an OGTT.16

Fig. 1.

Selection of the mothers for the original GDM screening study and the follow-up rates of their offspring in the current study.

Offspring Follow-up Study

Most offspring of mothers in the GDM screening study received subsequent health care at GHC, and childhood growth measurements were available in their outpatient medical charts. Because our hypotheses related to childhood obesity, offspring were eligible for this follow-up study only if their chart contained at least one height and weight measurement, recorded on the same day, on or after their fifth birthday. For eligible offspring, we obtained all height and weight measurements recorded before January 1, 1996, in the outpatient medical chart (unless from an emergency department visit). We excluded six children with conditions having a major impact on stature and/or adiposity (eg, cancer), and three offspring were stillborn.

Outcome Measures

We used body mass index (BMI) (weight [kilograms] divided by height [meters] squared) to assess fatness. Although BMI does not measure fatness directly, it is an acceptable surrogate measure of childhood fatness among indices derived from height and weight measurements.19,20 BMI in children is correlated with direct measures of adiposity,21 blood pressure,22 and serum lipid23 and insulin concentrations.24

BMI points (height and weight measurement recorded on the same day) were standardized for age and sex by conversion to a zscore. This standardization was required because children were measured at different ages and because BMI varies with age. To standardize BMI, we used z scores rather than percentiles, becausez scores are more normally distributed and becausez scores more clearly convey the magnitude of BMI difference between any two measurements at the extremes of the BMI distribution. The z score was calculated as (BMI − mean)/SD, where the mean and SD of BMI were from a reference population of the same age and sex as the subject. For points after 3 years of age, we used as a reference the combined data from National Health and Nutrition Examination Surveys I and II,25 and for points before 3 years of age, we used data from the Fels Longitudinal Study (S. Guo, personal communication, September 20, 1995). Means and SD values of BMI for specific ages (eg, 6.2 years) were found by linear interpolation between discrete ages (eg, 6 years and 7 years) given in the reference data.

For each subject, we calculated the average BMI z score between 5 and 10 years of age. For subjects with two or more BMI points between 5 and 10 years of age, we estimated the average BMIz score by interpolating data linearly between available points and extrapolating the first and last points out to the ends of the 5- to 10-year interval. The formula used for the average was therefore,i=1K+112(BMIz(ti)+BMIz(ti1))(titi1)/(tK+1t0),where the ages at which BMI points were measured weret1, … tK and where the endpoints of the time interval were t0 = 5 andtK+ 1 = 10. The formula can also be rewritten to show that it computes a weighted average with BMI points closely spaced in time receiving relatively less weight than widely spaced points. This weighting ensures that the average is not unduly influenced by multiple observations clustered close together in time.

Although there is no established BMI cut-point to define childhood obesity,26 subjects were classified as obese if, between 5 and 10 years of age, their average BMI z score was ≥1.036, which corresponds to the 85th percentile of a normal distribution. We also calculated the average BMI for age intervals before 5 years and the BMI at birth to demonstrate, using a consistent measure across ages, how offspring fatness changed from birth through age 10 years. Weight for height is an alternative surrogate measure of fatness in children, especially at younger ages. However, this measure could not be calculated for a number of subjects at the older ages because the National Center for Health Statistics weight-for-height charts do not use data for males taller than 145 cm or females taller than 137 cm.27 Birth–weight ratio was calculated as another surrogate measure of birth size, because the ratio provides a continuous measure of birth weight adjusted for gestational age and sex. The birth–weight ratio was calculated by dividing the offspring birth weight by the median birth weight for gestational age and sex (based on a reference population of non-Hispanic white newborns).28 Infants were considered large for gestational age if they had a birth–weight ratio ≥1.15, ie, a birth weight ≥115% of reference weight for gestational age and sex. This cut-point is equal to ∼4000 g for infants born at 40 weeks' gestation.

Predictor Measures

Because there were only five insulin-treated mothers, we excluded them from our primary analysis. The main comparison of interest was between the offspring of mothers with GDM (treated with diet alone) and the offspring of control mothers (those with a normal glucose screen value). Among all offspring with available data, we also examined the relationship between offspring obesity in childhood and four metabolic markers of GDM. Three markers (maternal screening plasma glucose, triglyceride, and glucose tolerance) are indirect measures of the maternal metabolic substrate available to the fetus. The fourth marker, serum insulin from the offspring cord blood, is an indirect measure of the fetal hyperinsulinemia induced by the increased transplacental transfer of maternal glucose.29,30 The OGTT results were summarized for each mother as the area under the OGTT curve, with a larger area indicating greater glucose intolerance. The laboratory measurement of all specimens and the calculation of the area under the OGTT curve were described previously.16,17

Maternal obesity before pregnancy and paternal obesity at offspring delivery were the covariates considered. Maternal prepregnant BMI was based on self-reported prepregnant weight and the measured height, both recorded at the first prenatal visit. Paternal BMI at offspring delivery was estimated from the available height and weight measurements in the father's GHC medical record. If paternal height was available in the medical record, then BMI points were calculated for the recorded paternal weight measurements. The paternal BMI on the day of offspring delivery was estimated by linear interpolation between paternal BMI points before and after the delivery date. Parent obesity was defined as a BMI ≥27.8 in fathers and ≥27.3 in mothers.31

Statistical Analysis

Rates of obesity and mean BMI z scores in GDM and control offspring were compared with χ2 and Wilcoxon rank sum tests, respectively. Offspring were divided into quartiles by the value for each of the four metabolic markers of interest. We tested the association between quartiles of each metabolic marker and offspring obesity using logistic regression with likelihood ratio tests. Multivariable logistic regression analyses of obesity rates also were performed, controlling for the effects of parental obesity.


Of the original 977 offspring, 524 (54%) met criteria for follow-up. Figure 1 shows follow-up rates by maternal GDM screening group. The majority of those not followed-up had disenrolled from GHC before 5 years of age (62%) or had no health care visits to GHC after 5 years of age despite being enrolled (16%). Fifty-one percent of the eligible offspring were boys, 94% were non-Hispanic whites, and 93% were born to married mothers. Subjects had a median of 2 BMI points recorded after 5 years of age (range, 1 to 17 points), with the most recent BMI point at a median age of 8.0 years. The age distribution of BMI points was similar in GDM and control offspring (data not shown). Table 1 describes the offspring and their parents. Twenty percent of the children were obese between 5 and 10 years of age, which is consistent with current US trends.32 The obesity rate was higher in the fathers than in the mothers.

Table 1.

Characteristics of Study Cohort

The offspring of diet-treated mothers with GDM tended to have lower obesity rates and BMI z scores than offspring of control mothers, but neither difference was statistically significant (Table2). Even when we used a higher BMI cut-point to define obesity (average BMI z score ≥1.645 or approximately the 95th percentile of BMI for age and sex), there still was no significant difference in obesity rates between GDM and control offspring (12.1% vs 11.7%; P = .93). Three of the five offspring of insulin-treated mothers were obese. When these five offspring were combined with the offspring of diet-treated mothers, there still was no increased risk of obesity (or no higher mean BMIz score) in offspring of mothers with GDM compared with offspring of controls (22% vs 24%; P = .75). The OGTT criteria for the diagnosis of GDM, which were established by the National Diabetes Data Group,9 are more stringent than those we used. When these stricter criteria were applied to our cohort, there were 37, rather than 58, offspring of diet-treated mothers with GDM. The rate of obesity in this group of 37 offspring of mothers with GDM was 27% and still was not significantly greater than the rate of 24% among controls (P = .70).

Table 2.

Comparison of Offspring BMI z Scores and Obesity Rates at Age 5 to 10 Years by Maternal GDM Screening Group and Parent Obesity Status

There was a significantly higher obesity rate in children whose mothers or fathers were obese (Table 2). Because parent obesity is a strong risk factor for childhood obesity,33 and because maternal obesity also is a risk factor for GDM,11 we evaluated all associations of childhood obesity and maternal GDM status while controlling for parent obesity. In these adjusted analyses, the risk of obesity was no higher in the offspring of mothers with GDM than in offspring of control mothers. Offspring of mothers with a normal glucose screen (plasma glucose <7.77 mmol/L [140 mg/dL]) had a higher rate of obesity than offspring of mothers with an abnormal screen (controls). After adjusting for parent obesity, however, the difference in childhood obesity rates between these two groups was not statistically significant.

We also compared the offspring obesity rates and BMI zscores before 5 years of age (Table 3). Offspring of mothers with GDM were larger at birth with a greater proportion classified as large for gestational age by birth–weight ratio (P = .02) or birth BMI (P = .06). Between 6 and 12 months of age, however, these differences not only disappeared, but there was a suggestion that offspring of mothers with GDM were leaner.

Table 3.

Comparison of GDM and Control Offspring BMI z Scores and Obesity Rates by Age*

Women in the higher quartiles of screening glucose level had offspring with lower obesity rates (Table 4). However, after adjustment for parent obesity, the differences in childhood obesity rates were no longer statistically significant. There were no significant differences in the rate of childhood obesity across quartiles of maternal glucose tolerance, maternal triglyceride, or cord serum insulin. When these four metabolic measures were each examined as continuous variables and correlated to offspring BMI zscores, there were no significant correlations. The five offspring with insulin-treated mothers were excluded from the analyses in Table 4, but the overall results were unchanged when these five cases were included.

Table 4.

Comparison of Offspring BMI z Score and Obesity Rates by Quartile of Metabolic Measures From Pregnancy and by Parental Obesity*

Our sample size was adequate to detect a clinically relevant association between maternal GDM and offspring obesity in childhood. Given the prevalence of obesity in the control group (24%) and the number of offspring available for follow-up (n = 58 for GDM; n = 257 for controls), our study had 85% power (α = 0.05) to detect a relative risk of 1.8 for obesity in the offspring of mothers with GDM.

We explored the possibility that differential follow-up in the GDM and control groups masked an association between GDM and offspring obesity (Table 5). In the GDM group, those offspring followed-up tended to have mothers with lower prepregnancy BMI and obesity rates. In contrast, for the control group, those offspring followed-up tended to have mothers with higher prepregnancy BMI and obesity rates. These differences could have biased our results toward the finding of no difference in obesity rates between offspring of control mothers and offspring of mothers with GDM, but when we controlled our analyses in Table 2 for maternal obesity, our conclusions were unchanged.

Table 5.

Comparison of GDM and Control Groups by Follow-up Status


We observed no increased risk of childhood obesity in offspring of mothers with mild, diet-treated GDM and found no association between the metabolic markers of GDM and childhood obesity. Our conclusions apply to the population we studied, namely, the offspring of medically insured, non-Hispanic white mothers with diet-treated GDM. The GDM prevalence in the HMO population from which the study cohort was derived16 was very similar to prevalence estimates in other populations of non-Hispanic whites. However, the cohort may have had less severe glucose intolerance at presentation or after diagnosis than in populations with poorer access to health care or higher rates of either GDM or maternal obesity.10,34 It is possible that intrauterine exposure to more severe maternal diabetes may increase the risk of childhood obesity. Although meaningful statistical comparisons cannot be made with our subgroup of five insulin-treated mothers, these five women represented a minority of the GDM cases identified in the population-based sample of 2019 women who were screened.

All previous studies of childhood obesity in the offspring of mothers with GDM also have included mothers with known IDDM12,14,15,35 or NIDDM.13 A Swedish study, the only to report results separately for offspring of mothers with GDM, showed no association between GDM and childhood obesity.35 The other studies, grouping all mothers with diabetes, reported an overall association between maternal diabetes during pregnancy and offspring obesity. In one report from the Northwestern Diabetes in Pregnancy Study, BMI in the offspring of mothers with diabetes was compared with that in a control group.15 Mean BMI values at an average age of 12 years were higher in the 88 offspring of mothers with diabetes than in the 80 control offspring (22.8 vs 20.3). This finding is difficult to interpret, because the offspring of mothers with GDM were not examined separately; the control group was not matched on markers of social status; and the BMI values were not adjusted for age, sex, stage of sexual maturity, or maternal BMI. Adjusting for maternal obesity is necessary to determine whether the offspring obesity risk is from the effects of an altered intrauterine environment or from obesity genes inherited from the mother. Effects from the intrauterine environment are potentially modifiable by interventions to improve glucose control during pregnancy.

Pettitt and colleagues, studying the Pima Indians in Arizona, compared obesity rates in a large cohort of offspring from diabetic and nondiabetic pregnancies.13 The risk of obesity (>140% of median weight for height) was two to three times higher in childhood and adolescence among the offspring of mothers with diabetes. This increased risk was present independent of both maternal obesity and birth weight.36 However, the high underlying genetic predisposition to obesity and diabetes in the Pima37 makes the findings from this population difficult to generalize, and these reports do not indicate what proportion of mothers with diabetes had known NIDDM before pregnancy.38

Our study focused on obesity in children between 5 and 10 years of age. Data from the previous studies suggest that it is not until after ∼5 years of age that the weights of offspring of diabetic mothers begin to differ from the growth reference14,39 or from controls.12 Thus, differences between offspring of mothers with GDM and offspring of control mothers may have been diminished in our study by averaging BMI over the ages of 5 to 10 years. However, when only those subjects with measurements between 8 and 10 years of age were compared (33 GDM vs 133 control offspring), there still was no significant difference in obesity rates (18% vs 19%;P = .93) or in mean BMI z score (0.38 versus 0.30; P = .80) at 8 to 10 years of age.

Our study provided an ideal control group. The screening criteria used on the mothers to establish GDM or control status were identical, and all mothers were enrolled at the same time. Furthermore, because the control mothers and mothers with GDM were all insured by the same HMO, this minimized differences in socioeconomic status and access to medical care that may be related to both glucose control in pregnancy and offspring obesity risk. Race and ethnicity, which also are important factors in both GDM risk10 and childhood obesity risk40 were not confounding variables in this study. We had incomplete follow-up of this cohort, but we were able to adjust our analyses for baseline differences in maternal obesity, and this adjustment did not affect our conclusions.

Our findings indicate that mothers with mild, diet-treated GDM do not have metabolic alterations that affect the intrauterine environment sufficiently to increase the risk of childhood obesity in their offspring. It is possible that any effects of GDM on the intrauterine environment that increase childhood obesity risk are obscured by the greater impacts of diet and physical activity patterns during childhood. Our results, together with those of previous studies, suggest that the risk of childhood obesity in offspring from diabetic pregnancies may depend on the form and severity of maternal diabetes. Future studies must examine separately mothers with IDDM, NIDDM, and GDM. To delineate further the possible relationship between diabetes during pregnancy and offspring obesity, these studies must examine prospectively fat and carbohydrate metabolism and body fat distribution both in pregnant mothers and in their offspring.


This work was supported by the Generalist Physician Faculty Scholars Award from the Robert Wood Johnson Foundation, Princeton, NJ (R.W.), and by Grant DK35816 to the University of Washington Clinical Nutrition Research Unit from the National Institutes of Health (R.K.).

We thank Edward H. Wagner, MD, MPH, for facilitating this research at the Center for Health Studies at Group Health Cooperative of Puget Sound, Seattle, WA; Richard L. Furman for careful abstraction of Group Health Cooperative medical records; and Vicki Livengood for her assistance with the preparation of this manuscript.


    • Received July 16, 1997.
    • Accepted October 14, 1997.
  • Reprint requests to (R.C.W.) Children's Hospital Medical Center, Division of General and Community Pediatrics, CH-1S, 3333 Burnet Ave, Cincinnati, OH 45229-3039.

  • This paper was presented at the meeting of the North American Association for the Study of Obesity, Breckenridge, CO, October 13, 1996.

insulin-dependent diabetes mellitus
gestational diabetes mellitus
noninsulin-dependent diabetes mellitus
Group Health Cooperative of Puget Sound
health maintenance organization
oral glucose tolerance test
body mass index