Determinants of Mild Gestational Hyperglycemia and Gestational Diabetes Mellitus in a Large Dutch Multiethnic Cohort

Consecutive pregnant women (1,059, over a study period of 30 months [1998–2001]) were recruited from the outpatient clinic of obstetrics and gynecology of our hospital. The patients were referred due to a history of delivery by caesarean section (30.6%), high-risk pregnancy (10.5%), stillborn fetus (4.4%), or multiple pregnancy (6.5%), or they were referred for (preexistent) hypertension (6.4%), maternal age (5.6%), treatment of infertility (3.6%), suspicion of GDM (3.5%), or other obstetrical pathology (28.9%). Diabetic patients (n = 11) were excluded from the protocol. Informed consent was obtained, and the study was approved by the hospital’s ethics committee.

Obstetric history, family history of diabetes, and weight before pregnancy were obtained from all patients. As indicators of pregnancy outcome, we studied birth weight, expressed as a percentile specific for sex and duration of gestation (5), and Apgar scores at 1 and 5 min after birth.

All women had a 50-g GCT at 16–33 weeks of gestation, performed in the morning after a 12-h overnight fast (6). Fasting glucose was measured in venous whole blood, with a capillary whole blood glucose measurement 1 h later. If the latter yielded a value of ≥7.8 mmol/l, a 3-h 100-g OGTT followed within the next 2 weeks. Fasting glucose was measured in venous whole blood, followed by a capillary whole-blood glucose measurement 1, 2, and 3 h after 100 g glucose. GDM was diagnosed when two or more glucose values equaled or exceded 4.6, 9.6, 8.2, and 7.3 mmol/l, respectively. We converted the 1997 American Diabetes Association criteria (7) according to the data published by Alberti and Skrabalo (8).

A standard radioimmunoassay kit method was used to measure fasting serum C-peptide (Immulite C-Peptide; EURO/DPC, Llanberis, U.K.; overall coefficients of variation were 14 and 8% at 0.7 and 1.7 nmol/l, respectively) (9). Whole- blood glucose was determined by a hexokinase method using a glucose analyser (EBIO model 6666; Eppendorf, Hamburg, Germany).

Ethnicity was assigned on the basis of three criteria: country of birth of the individual, the mother, and the father. It was coded according to the definitions used by the O+S, the Amsterdam bureau of social research and statistics (10). Ethnicity was categorized as Caucasian (512 [48.4%]), sub-Saharan African (174 [16.5%]), Northern African (120 [11.3%]), Armenian (58 [5.4%]), Asian (55 [5.2%]), and others (140 [13.2%]).

BMI (kg/m²) before pregnancy was calculated by using the most recent self-reported weight before conception. Obesity was defined as a BMI ≥27 kg/m² (11). Pregnancy-induced hypertension was defined as an increase in diastolic blood pressure to >90 mmHg (measured using a Korotkoff 5; Speidel & Keller, Jungingen, Germany) in the second half of pregnancy (12). Macrosomia was defined as a birth weight exceeding the 90th percentile (13). We defined a neonate to be at risk when the Apgar score at 5 min was ≤6 (14). We introduced the term MGH to describe women who had a positive GCT followed by a negative OGTT.

In our analysis, the following variables were included: maternal and gestational age, prepregnancy BMI and obesity, ethnicity (Caucasian versus non-Caucasian), family history of diabetes, parity, history of spontaneous abortion and of fetal loss, prepregnancy/pregnancy-induced hypertension, birth weight (percentiles), Apgar score, C-peptide, and glucose. Variables with nonnormal distribution (e.g., C-peptide) were log-transformed before statistical analysis. To compare variables among control, MGH, and GDM women, data were analyzed using one-way analysis of variance and the Kruskal-Wallis test where appropriate. The Bonferroni t method or the Mann-Whitney U test were used as post hoc tests. To compare variables such as C-peptide and glucose among the three diagnostic groups, which were matched by prepregnancy BMI, we used analysis of covariance, with prepregnancy BMI as covariate.

We used a multiple logistic regression model to identify independent determinants of MGH and GDM. The value for entry in the logistic regression model was P = 0.10. The following variables were considered as possible determinants: maternal age, ethnicity, parity, pregnancy-induced hypertension, and prepregnancy BMI.

Two-sided P values are reported, with P = 0.05 taken as the level of statistical significance. All statistical procedures were performed using SPSS version 9.0 for Windows (SPSS, Chicago, IL).

We used bivariate analysis to determine reference ranges of combined fasting glucose and C-peptide levels, i.e., the 95% isodensity ellipse of the distribution, and to calculate the absolute correlation coefficient between the mentioned variables (15). The calculation was according to Tatsuoka (16), using the software package EVAL-kit (CKCHL Twee Steden Ziekenhuis, Tilburg, the Netherlands).

A total of 1,059 eligible women gave informed consent to participate in the study. Four patients suffered from Graves disease. Twenty-two patients were unable or unwilling to follow the protocol, i.e., glucose concentration measured too late, one or several glucose measurements missed, or refusal to finish the tolerance test. Eleven women with newly detected type 2 diabetes could not enter the study, leaving 1,022 patients.

GDM was diagnosed in 71 women (6.9%) and MGH in 138 women (13.5%). From the remaining 813 women with a negative GCT, 490 were randomly selected to serve as control subjects. Therefore, the final study population consisted of 699 women (66%) who entered the study between 16 and 33 weeks of gestation.

Clinical and laboratory characteristics of control, MGH, and GDM subjects are outlined in Table 1. The five mentioned non-Caucasian subgroups were too small to permit any valid mutual comparisons. Therefore, these women were combined as non-Caucasians. The relatively large proportion of non-Caucasians in the study group reflects the population constitution in the area (17). Table 1 also shows the results of multiple-comparison procedures examining possible determinants of MGH and GDM. GDM women, compared with control subjects, had markedly higher (P < 0.05) maternal age, prepregnancy BMI, rate of non-Caucasian origin, parity, and pregnancy-induced hypertension, whereas only maternal age and prepregnancy BMI were increased (P < 0.05) in MGH versus control women. The prevalence of family history of diabetes was highest but not significantly different in GDM women compared with control and MGH women. In contrast to MGH women, GDM women were at significantly increased risk for delivery of a macrosomic infant (13/138 [9.4%] vs. 25/71 [35.2%]) and/or an infant with a low Apgar score at 5 min.

Determinants that were (close to) significant (P < 0.10) in the univariate analysis were entered into a multiple logistic regression analysis. Table 2 shows the results of multivariate analyses with MGH and GDM as the dependent variable. Maternal age, ethnicity, and prepregnancy BMI were significantly associated with GDM and explained 19% of its variance. Maternal age and prepregnancy BMI were significantly associated with MGH and explained 7% of its variance.

Fasting glucose, glucose at 60 min during GCT, and fasting C-peptide were significantly increased (P < 0.001) in MGH women by 8, 39, and 43%, respectively, compared with control subjects (Table 1). In parallel, fasting glucose, glucose at 60 min during GCT, and fasting C-peptide were significantly increased (P < 0.001) in GDM women by 31, 61, and 67%, respectively, compared with control subjects. Similar results were obtained when the difference in prepregnancy BMI was adjusted for (data not shown). Figure 1 displays the course of fasting C-peptide level during gestational age among the three diagnostic groups. It seems that fasting C-peptide level of control subjects becomes markedly increased during gestation and that for all pregnant women who were at 20–33 weeks of gestation, MGH induced a significant (21%) upward displacement of the C-peptide curve as compared with control subjects. The highest level of fasting C-peptide (∼1 nmol/l) with the lowest degree of increase during pregnancy was seen in GDM women.

The results of bivariate statistical analyses for the combination of fasting glucose and C-peptide are displayed in Table 3. Women with significantly outlying glucose or C-peptide values were excluded, resulting in omission of these data for three control and two MGH women. The data provided evidence of a positive correlation between fasting glucose and C-peptide for the control subjects (r = 0.39 [95% CI 0.31–0.46]), resulting in a slim reference ellipse (Fig. 2A). A similar result was seen for MGH women (r = 0.38 [0.23–0.52]) (Fig. 2B). The correlation was nearly lost in GDM women (0.14 [−0.09 to 0.36]), with the result that the shape of the reference ellipse approached a circle (Fig. 2C). In addition, the fasting C-peptide–to–glucose ratio, i.e., the slope of the major axis of the ellipse (the standard principal component), was significantly reduced (P < 0.001) in GDM women compared with control subjects (0.41 ± 0.25 and 0.70 ± 0.20, repectively), whereas the ratio was similar (P = 0.17) in both MGH and control women (0.73 ± 0.23 and 0.70 ± 0.20, respectively). After including the five outliers, similar results were seen for the bivariate statistical analyses (data not shown).

This study was undertaken to characterize determinants of MGH and GDM and to assess the extent of correlation between glucose and C-peptide in control, MGH, and GDM women.

Considering the high fertility rate in all immigrant groups, the ethnic composition (Caucasian/non-Caucasian) in both the control and MGH women (52/48% and 49/51%, respectively) was quite similar to the female population of the four town boroughs in close proximity to the hospital (57/43%). The observed increase of the C-peptide level during pregnancy is well documented (18). To the best of our knowledge, there are no other published data specifically addressing the association between fasting glucose and C-peptide from a cohort of pregnant women.

Overall C-peptide and glucose concentrations were significantly elevated in MGH compared with control women. However, both levels were lower than those of the GDM women. Furthermore, our results suggested that C-peptide levels in MGH women, in contrast to GDM women, parallelled those in control subjects during pregnancy. Both the correlation coefficient between fasting glucose and C-peptide and the fasting C-peptide–to–glucose ratio were found to be similar in MGH and control women. These findings suggest a near-normal pattern of glucose-triggered insulin release in MGH. The present study identified two determinants of MGH: maternal age and BMI. However, there was no association between MGH and ethnicity.

GDM affects 3–4% of pregnant women (19,20). In this study, we found a prevalence of 6.9%, likely due to the high rate of immigrants in the Amsterdam-East region (17).

Epidemiological studies have shown that the risk of adverse birth outcomes is strongly related to GDM (20,21,22). The results of the present study, i.e., fetal macrosomia and low Apgar scores at 5 min, are in keeping with the results reported in those studies. Our study largely confirms the results of previous studies addressing determinants of GDM. Maternal age (23,24,25), ethnicity (23,24,25), and prepregnancy BMI (23,24) were positively associated with GDM but not pregnancy-induced or prepregnancy hypertension (26). This study showed significantly higher serum C-peptide levels in GDM women than in MGH women and control subjects. In GDM women, the correlation between fasting glucose and C-peptide was almost lost and, in addition, the fasting C-peptide–to–glucose ratio was two times less than in control subjects. These findings suggest a higher level of insulin resistance in GDM compared with MGH women and are compatible with the concept that glucose intolerance in GDM is (partly) due to an impaired ability of β-cells to further increase insulin secretion in response to glucose (27,28). The fact that ethnicity is the only factor distinguishing MGH from GDM suggests that genetic or lifestyle factors are involved in this type of β-cell dysfunction.

Finally, it is noteworthy that a population-based survey performed in one of the four above-mentioned town boroughs showed that all immigrant subjects had a significantly higher risk for type 2 diabetes than Caucasian subjects and that a significant association was found between non-Caucasian origin and the occurrence of GDM (17).

There are a number of caveats to the data presented here. Firstly, patient selection in the current study design was based on referral to an outpatient clinic of obstetrics and gynecology. Because of the selection bias that may have been introduced by this, we feel caution is required in extrapolating data of the control group to the general population. In control women, however, ethnicity was quite similar when compared with the female inhabitants of the four town boroughs in close proximity to the hospital. Also, in our study, the established ranges of BMI, fasting glucose, and C-peptide in control women support the data of previous reports that studied relatively small random samples of the general population (29,30). Therefore, the data presented here offer some evidence in favor of representativeness. Secondly, the causal relation between the significant determinants of MGH and GDM cannot be proven using cross-sectional data. The close association shown in our study suggests a causal relation that requires further study.

In summary, MGH is characterized both by insulin resistance and a near-normal pattern of glucose-triggered insulin release. Of the well-known independent determinants of GDM, only maternal age and prepregnancy BMI were associated with MGH. Thus, it appears that additional factors promoting loss of β-cell function distinguish MGH from GDM. One of these factors appears to be ethnicity.