Gestational diabetes and pre-pregnancy overweight: Possible factors involved in newborn macrosomia


Dr Pablo R. Olmos, Department of Nutrition, Diabetes and Metabolism, College of Medicine, Pontificia Universidad Católica de Chile, Alameda 340, Santiago 6513492, Chile. Email:


Aim:  Good glycemic control in gestational diabetes mellitus (GDM) seems not to be enough to prevent macrosomia (large-for-gestational-age newborns). In GDM pregnancies we studied the effects of glycemic control (as glycosylated hemoglobin [HbA1c]), pre-pregnancy body mass index (PP-BMI) and gestational weight gain per week (GWG-W) on the frequency of macrosomia.

Methods:  We studied 251 GDM pregnancies, divided into two groups: PP-BMI < 25.0 kg/m2 (the non-overweight group; n = 125), and PP-BMI ≥ 25.0 kg/m2 (the overweight group; n = 126). A newborn weight Z-score > 1.28 was considered large-for-gestational-age. Statistical analysis was carried out using the Student's t-test and χ2-test, receiver–operator characteristic curves and linear and binary logistic regressions.

Results:  Prevalence of macrosomia was 14.9% among GDM (n = 202/251, 88.4%) with good glycemic control (mean HbA1c < 6.0%), and 28.1% in those with mean HbA1c ≥ 6.0% (n = 49/251, P < 0.025). Macrosomia rates were 10.4% in the non-overweight group and 24.6% in the overweight group (P = 0.00308), notwithstanding both having similar mean HbA1c (5.48 ± 0.065 and 5.65 ± 0.079%, P = 0.269), and similar GWG-W (0.292 ± 0.017 and 0.240 ± 0.021 kg/week, P = 0.077). Binary logistic regressions showed that PP-BMI (P = 0.012) and mean HbA1c (P = 0.048), but not GWG-W (P = 0.477), explained macrosomia.

Conclusions:  Good glycemic control in GDM patients was not enough to reduce macrosomia to acceptable limits (<10% of newborns). PP-BMI and mean HbA1c (but not GWG-W) were significant predictors of macrosomia. Thus, without ceasing in our efforts to improve glycemic control during GDM pregnancies, patients with overweight/obesity need to be treated prior to becoming pregnant.


There is a worldwide consensus that delivery of a macrosomic or large-for-gestational-age (LGA) infant is associated with increased frequencies of prolonged labor, operative delivery, shoulder dystocia and brachial plexus trauma.1 In the particular case of the macrosomia that is due to gestational diabetes mellitus (GDM), maternal hyperglycemia – and its consequence, fetal hyperinsulinemia – are positively correlated to neonatal excess body mass.2 However, tight glucose control seems not to be enough to prevent macrosomia in GDM, as other variables have emerged as independent factors of excessive fetal growth, particularly maternal overweight and obesity (body mass index [BMI] of 25 or greater).3

In fact, in the last 2 decades, evidence has accumulated showing that maternal obesity is a risk factor for adverse pregnancy outcomes in general (i.e., even in the absence of GDM), and LGA infants in particular.4,5

In 2004, a large review of maternal and neonatal data in the USA6 showed that the statistical association between GDM and LGA infants strengthened as pregravid BMI increased, thus suggesting that maternal overweight was an independent risk factor for LGA newborns among those patients. One year later, Ricart et al.7 showed that a pregravid BMI above 26.1 kg/m2 was a risk factor for LGA that not only was independent from the presence of GDM, but was also a stronger predictor for LGA than the degree of hyperglycemia as determined by the results of the oral glucose tolerance test (OGGT).

Apart from pre-pregnancy overweight and obesity, little is known about the impact that weight gain during pregnancy might have on excessive fetal growth in the particular case of gestational diabetes, as recent studies have shown that a single recommendation of weight gain is not feasible based on current data. In fact, although the subject seemed to be settled in 1990 when the US Institute of Medicine established their gestational weight gain recommendations,8 a 2009 review on outcomes of gestational weight gain, despite finding a positive correlation between weight gain and risk of LGA infants,9 also revealed that the quality of published data was inadequate for providing a single recommendation of gestational weight gain to all cases – in agreement with the conclusions of a contemporary review of 189 studies.10 In other words, adjusting gestational weight gain to pre-pregnancy maternal nutritional status is not enough, as other confounding variables and comorbidities need to be taken into consideration – one of which is GDM.

In this context, the aim of our study was to establish in women with GDM whether, in addition to glycemic control, pregravid overweight and/or pregnancy weight gain have an impact on the newborn's weight and the frequency of LGA.

Materials and Methods

This study on GDM included 251 pregnancies (in 245 mothers), all of which had been referred by their obstetrician to our diabetes and pregnancy team. The ‘Healthcare Network’ of our College of Medicine includes six Obstetrics and Gynecology outpatient clinics in Santiago, where faculty and staff carry out the evaluation and follow up of pregnant mothers, who in due course are admitted for delivery at the teaching hospital of our institution. We assumed a 4% rate of GDM, which is the arithmetic mean of the figures from two references, that is, 4.1% published in 2004 by Catalano et al.6 and 4.0% (central value of a 2–6% range) given in the year 2010 by Galtier.11 With this consensus figure of 4.0%, plus an average of 2200 deliveries per year in our hospital, we expected to recruit a maximum of 88 new cases of GDM per year, that is, 968 GDM pregnancies between 1998 and 2009, the period during which the recruitment and follow up of our patients was made. A total of 251 GDM pregnancies were referred to our diabetes and pregnancy team, that is, 26% of the expected patients. During these 11 years, however, the GDM referral rate has increased steadily (up to 80% of the expected number of cases in the year 2008). Our data does not include multiple pregnancies.

The diagnosis of GDM was confirmed by a World Health Organization OGGT, using a 75-g oral glucose load, with basal and 2-h venous plasma glucose measurements.12,13 Referral of GDM mothers was done after the 24th week, which is the date at which the OGGT is carried out in all pregnancies in Chile.

The diet consisted of 25–30 kilocalories per kg of ideal pre-pregnancy bodyweight for women with normal bodyweight, aiming to an increase of 11–12 kg during pregnancy. The kilocalories allowance was moderately reduced in obese women (i.e., a pre-pregnancy BMI ≥ 30 kg/m2) in order to limit the weight gain to 6–7 kg. Approximately 40–50% of total calories were carbohydrates, excluding saccharose, grapes and ‘diet’, ‘light’ or ‘zero’ products (sweet carbonated beverages, ice-creams and jams). Each of the GDM mothers received an individualized diet both verbally and as a detailed letter written and printed during the outpatient visit by the endocrinologist. Compliance to the diet was assessed in each medical visit, and appropriate counseling was given in case errors were detected.

Recommended results for glucose (mg/dL) self-monitoring were as follows: before breakfast = 70–90; 1 h after breakfast = 90–120; before the midday meal = 90–105; 1 h after the midday meal = 90–120; before the evening meal = 90–105; and 1 h after the evening meal = 90–120.14 Whenever these parameters could not be met by diet and self-monitoring of blood glucose, patients were admitted to the institutional hospital, where they were started on intensified insulin therapy (IIT).

In general, ‘glycemic control’– also known as ‘quality of glycemic control’– is defined as the extent to which serum glucose of a given patient falls within recommendations. The variable that gives the best measure of glycemic control is the glycosylated hemoglobin (HbA1c). The HbA1c was measured every 2 weeks throughout, by means of high-performance liquid chromatography, using a Variant-II HPLC system, with a normal range of 4.27–6.07%, and a standard deviation of 0.45%. In this manuscript, ‘mean HbA1c’ is the arithmetic mean of all measurements of HbA1c that have been done in a given patient during the entire pregnancy.

Another variable reflecting glucose control is the mean blood glucose, which is the arithmetic mean of all capillary glucose measurements of a given patient from diagnosis of GDM to delivery.

IIT consisted of two injections (before breakfast and at bedtime) of neutral protamine hagedorn (NPH) human insulin (Humulin-N or Insuman-N), plus three injections of regular (rapid-acting) insulin (Humulin-R or Insuman-R). All patients attended follow-up appointments by an endocrinologist and an obstetrician every 1–2 weeks.

We used the 2004 percentile curves for fetal weight of the Chilean Ministry of Health.15 The fetuses and newborn infants whose weights were between the 10th and 90th percentiles were considered adequate for gestational age, those below the 10th percentile were small-for-gestational-age, and those above the 90th percentile (i.e., weight Z-Score > 1.28) were classified as LGA.

Gestational weight gain of the mother (GWG) was calculated by subtracting the pre-pregnancy bodyweight from the antepartum bodyweight, whereas the gestational weight gain per week (GWG-W, in kg/week) was calculated by dividing the GWG by the duration of pregnancy.

Unless stated otherwise, results are reported as the mean ± one standard error of the mean. The statistics package used was pasw Statistics 18, version 18.0.2, dated 18 April 2010 (with WinWrapBasic, Copyright 1993–2007 Polar Engineering and Consulting ( Comparisons of means were carried out using the Student's t-test. Categorical variables were compared using the χ2-test. Linear regressions were carried out, calculating the ‘R’ coefficient and a P-value. Binary logistic regression with forward-conditional method was used to assess the impact of multiple variables on newborn macrosomia. Akaike information criterion was used to select the model with the best combination of variables. In order to determine the capacity of a continuous variable for discriminating LGA from non-LGA newborns, receiver–operator characteristic (ROC) curves were constructed to find the best possible trade-off between sensitivity and ‘1-specificity’ (false positive) rates. The area under the ROC curve was also calculated.

A P-value below 0.05 was considered significant throughout. Bonferroni correction was used as needed for multiple comparisons.


In 251 pregnancies with GDM, 17.7% were LGA newborns. Prevalence of macrosomia was 14.9% among GDM (n = 202/251, 88.4%) pregnancies having good glycemic control (mean HbA1c < 6.0%), and 28.1% in those with mean HbA1c ≥ 6.0% (n = 49/251, P < 0.025).

Figure 1a shows the correlation between pre-pregnancy BMI and the Z-Score of the newborn weight in all 251 pregnancies with GDM. The data fit a linear regression function with a positive slope. Figure 1b shows the ROC curve for pre-pregnancy BMI as a predictor of macrosomia in the newborn. A cut-off point of 24.5 kg/m2 for pre-pregnancy BMI had a sensitivity of 0.83, a ‘1-specificity’ (false positive rate) of 0.53, and an area under the ROC curve of 0.703.

Figure 1.

(a) Correlation and linear regression between pre-pregnancy body mass index (BMI) and the Z-Score of the newborn weight in all 251 pregnancies with gestational diabetes mellitus. r = 0.2323, and P = 0.002. (b) Receiver–operator characteristic (ROC) curve for pre-pregnancy BMI used for prediction of macrosomia in the newborn, defined as Z-Score > 1.28 (i.e. >90th percentile). A pre-pregnancy BMI with a cut-off point of 24.5 kg/m2 had a sensitivity of 0.83, a ‘1-specificity’ (false positive rate) of 0.53, and an area under the ROC curve of 0.703.

Considering that the aforementioned cut-off point was very similar to the upper limit of the normal BMI in adults (25.0 kg/m2), we decided to divide the cohort of 251 GDM pregnancies into two groups (Table 1): the pre-pregnancy BMI < 25 kg/m2 (the non-overweight [NOW] group; n = 125) and the pre-pregnancy BMI ≥ 25 kg/m2 (the overweight [OW] group; n = 126). The OW group had a rate of macrosomia of 24.6%, which was 2.5 times higher than that in the NOW group (10.4%, P = 0.00308), despite having a similar mean HbA1c (P = 0.209) and similar GWG-W (P = 0.077).

Table 1.  Clinical profile (mean ± standard error of the mean) of both patients and newborns divided into two groups according to the pre-pregnancy weight status of the mother
(n = 125)(n = 126)
  • *

    Statistically significant: P < 0.003125 after Bonferroni correction has been applied. BMI, body mass index; GWG-W, gestational weight gain per week (kg); HbA1c, glycosylated hemoglobin; LGA, large-for-gestational-age (>90th centile = Z-Score > 1.28); NOW, non-overweight; OW, overweight.

Age (years)32.7 ± 0.6132.8 ± 0.43Student's t0.850
First visit (weeks)28.9 ± 0.6527.9 ± 0.59Student's t0.251
Pre-pregnancy BMI (kg/m2)22.6 ± 0.1530.0 ± 0.37Student's t
Mean HbA1c (%)5.48 ± 0.0655.65 ± 0.079Student's t0.269
Insulin therapy13 (10.4%)36 (28.6%)χ20.00028*
GWG-W (kg/week)0.292 ± 0.0170.240 ± 0.021Student's t0.077
Gestational age at delivery [weeks]37.9 ± 0.1637.5 ± 0.19Student's t0.163
Cesarean section (n[%])35 (28.0%)43 (34.1%)χ20.2943
Perinatal mortality (n[%])00χ21.00
Malformations (n[%])3 (2.4%)2 (1.58%)χ20.6450
Newborn weight >4000 g (n[%])6 (4.8%)15 (11.9%)χ20.0420
LGA newborn (n[%])13 (10.4%)31 (24.6%)χ20.00308*
Z-Score of the newborn weight0.17 ± 0.110.61 ± 0.14Student's t0.022
Hypoglycemia (n[%])2 (1.6%)2 (1.6%)χ21.00
Hyperbilirubinemia (n[%])8 (6.4%)16 (12.7%)χ20.08979
Admitted, intensive care (n[%])12 (9.6%)28 (22.2%)χ20.0063

In Table 2, binary logistic regression, limited to the NOW pregnancies only, shows that macrosomia is explained just by mean HbA1c, leaving out of the model both GWG-W and mean blood glucose.

Table 2.  Results in non-overweight pregnancies (n = 125) of binary logistic regression (method = ‘forward-conditional’) between newborn weight Z-Score > 1.28 and three covariables: mean HbA1c, GWG-W and MBG
 P-valueOdds ratio95% confidence interval
  • Odds ratio = e[B].

  • ‡Out of the three variables introduced, GWG-W and MBG were eliminated from the model on the first stage of the forward-conditional method of the binary logistic regression. GWG-W, gestational weight gain per week; HbA1c, glycosylated hemoglobin; MBG, mean blood glucose.

Mean HbA1c0.03618.2141.203–275.8

In Table 3, binary logistic regression, restricted only to OW pregnancies, showed that macrosomia is explained just by mean HbA1c, thus leaving out of the model both GWG-W and mean blood glucose.

Table 3.  Results in overweight pregnancies (n = 126) of binary logistic regression (method = ‘forward-conditional’) between newborn weight Z-Score > 1.28 and three covariables: mean HbA1c, GWG-W and MBG
 P-valueOdds ratio95% confidence interval
  • Odds ratio = e[B].

  • ‡Out of the three variables introduced, GWG-W and MBG were eliminated from the model on the first stage of the forward-conditional method of the binary logistic regression. GWG-W, gestational weight gain per week; HbA1c, glycosylated hemoglobin; MBG, mean blood glucose.

Mean HbA1c0.0303.2011.122–9.131

Binary logistic regression (Table 4) applied to the whole cohort of 251 GDM pregnancies, showed that both pre-pregnancy BMI (P = 0.012) and mean HbA1c (P = 0.048) explained macrosomia. The third variable, GWG-W (P = 0.477) had no significant impact.

Table 4.  Results in all pregnancies (n = 251) of binary logistic regression (method = ‘forward-conditional’) between newborn weight Z-Score > 1.28 and three covariables: pre-pregnancy BMI, mean HbA1c, and GWG-W
 P-valueOdds ratio95% confidence interval
  • Odds ratio = e[B].

  • ‡Out of the three variables introduced, only GWG-W was eliminated from the model on the first stage of the forward-conditional method of the binary logistic regression. BMI, body mass index; GWG-W, gestational weight gain per week; HbA1c, glycosylated hemoglobin; MBG, mean blood glucose.

Pre-pregnancy-BMI 0.0121.1501.031–1.282
Mean-HbA1c 0.0482.7141.031–7.304
Constant 0.0000.000


We have reported the study of the follow up and outcomes of 251 pregnancies treated by us for gestational diabetes, where the overall neonate LGA rate (17.7%) was related not only to glycemic control – defined as mean HbA1c – but also to maternal pre-pregnancy BMI. In fact, when analyzing the latter by means of a ROC curve, we found that a pre-pregnancy BMI equal or above 24.5 kg/m2 had a sensitivity (true positive rate) of 0.83 (i.e. 83%) in predicting LGA newborns, thus suggesting that maternal overweight – not even obesity – was enough to accelerate fetal growth. We say this because the accepted cut-off point for overweight in adults is 25.0 kg/m2, a figure that is very near to the one we obtained in the ROC curve.

Therefore, our results have also shown that among GDM pregnancies there were two different groups of patients according to whether the pre-pregnancy BMI was below or equal to/above 25.0 kg/m2. These two groups have proved to be different with respect to the frequencies of both macrosomia and insulin treatment during pregnancy.

The OW pregnancies had significantly higher frequency of LGA newborns than their NOW counterparts. At first sight, this difference could be assigned to the tendency of overweight and obese mothers to have more severe insulin resistance and hyperglycemia than mothers of normal pre-pregnancy weight – thus leading eventually to more severe hyperglycemia – as cytokines produced by the adipose tissues have been shown to increase insulin resistance. One of these is tumor necrosis factor-alpha (TNF-α), a molecule that induces serine phosphorylation in the insulin receptor substrate-1, thus altering the insulin signaling within muscular and adipose cells.16 The combined effects of ‘adipose cytokines’– TNF-α, interleukin-6 and leptin – would enhance insulin resistance in obese mothers more than in their non-overweight counterparts. This phenomenon could explain why, in our study, glucose control was more difficult in the OW group, as evidenced by the 2.77-fold higher frequency of insulin treatment compared to the NOW group. Unexpectedly, however, this mechanism could not explain the higher rate of macrosomia in OW compared to NOW pregnancies, as the mean HbA1c was similar in both groups.

Thus, the high frequency of LGA in OW could not be explained only by the classical pathway of maternal hyperglycemia–fetal hyperglycemia–fetal hyperinsulinemia.17 We hypothesize that, in addition to glucose control as reflected by the HbA1c, a second variable, maternal pre-pregnancy overweight, has somehow a role enhancing the transport from the OW mother to the fetus of nutrients other than glucose, that is, amino acids and/or lipids.

It has been demonstrated that placental leptin enhances mother-to-fetus amino acid transport through the syncytiotrophoblast (STB).17 As leptin in adults is produced by the adipose tissue,18 one may be tempted to propose that maternal amino acids could be responsible for the infant macrosomia that was observed in the OW group. However, others have shown that there is no correlation between maternal leptin levels and birthweight of the offspring.18,19

Therefore, only the lipids remain as the nutrients potentially responsible for fetal overgrowth. Physiologically, placental triglyceride transport during the third trimester is enhanced by maternal hypertriglyceridemia, which in turn is due to two mechanisms: the decreased catabolism of very-low-density lipoproteins (VLDL), plus the enhanced lipolysis, that in turn increases the substrate fatty acids and glycerol flow towards the hepatic synthesis of VLDL. Both pathways can be enhanced by maternal-overweight-related insulin resistance. The net result is a sharp increase in the flow of maternal VLDL towards the placenta.20–24

This hypothetical ‘lipid overflow’ could certainly explain the association that we have found between pre-gestational overweight and LGA infants in GDM women notwithstanding near-optimal glycemic control, as their insulin resistance was certainly made worse by adiposity. In this context, Schaefer-Graf et al.22 have shown that in GDM, maternal plasma triglycerides and fatty acids, measured in the third trimester, were independent predictors of LGA infants. Whether maternal excess of plasma triglycerides could be secondary to pre-pregnancy overweight is very likely, but it remains to be established.

Before closing the discussion, it is advisable to provide a word of caution on ruling out the variable GWG (Tables 2,3,4). This is because the health-care team in charge of these 251 pregnancies has made every effort to keep maternal weight gain within narrow limits. Thus, we may have unintentionally reduced the statistical impact of this variable on the regression models.

Despite our efforts towards tight control of blood glucose in gestational diabetic patients, the LGA rate stayed above what is regarded to be an acceptable limit (i.e., 10% or less). Our results suggest that, in order to further reduce the incidence of LGA infants in GDM, in addition to tight glycemic control during pregnancy in all patients, counseling and treatment for overweight and obese women is needed,25 encouraging them to aim for a BMI below 25.0 kg/m2 before becoming pregnant.

We conclude that in Chilean women with GDM, both pre-pregnancy BMI and mean HbA1c (but not GWG-W) are significant predictors of newborn macrosomia. Good glycemic control in GDM pregnancies was not enough to reduce macrosomia to acceptable (i.e. <10%) limits, as pre-pregnancy overweight/obesity (pre-pregnancy BMI > 24.5 kg/m2) remained as a variable that should be controlled in the pre-gestational period.26


The authors are grateful to Mrs Carolina Torres for her excellent secretarial support.