Introduction

Gestational diabetes mellitus (GDM) is defined as carbohydrate intolerance with onset or first recognition during pregnancy [1]. It increases the risk of adverse complications for both mother and child [1], a risk that can be reduced with appropriate diagnosis and treatment [2, 3]. In 1998, the Toronto Tri-Hospital Gestational Diabetes Project [4] reported an increased rate of pre-eclampsia, Caesarean section and macrosomia in women with untreated borderline GDM who fulfilled the more stringent Carpenter and Coustan criteria [5], but not the National Diabetes Data Group (NDDG) criteria [6]. In 2000, the American Diabetes Association (ADA) shifted from NDDG to Carpenter and Coustan thresholds [7], and with these criteria, GDM prevalence increased by an average of 50% in the multi-ethnic American population [8–10]. However, the predictive ability of these criteria in terms of pregnancy outcome has not been fully investigated in non-American populations. As the prevalence of macrosomia and pregnancy-induced hypertension (PIH)/pre-eclampsia is lower in Spanish women than in their Canadian counterparts [4, 11, 12], even if Spanish women with GDM according to ADA but not NDDG criteria had relative risks similar to those of Canadian women, their absolute risks would be lower. Thus, before deciding whether or not to adopt new criteria in the Spanish population, we evaluated the potential impact of a change from NDDG to ADA criteria on GDM prevalence, and the morbidity in women diagnosed as having GDM according to ADA but not NDDG criteria.

Subjects, materials and methods

This prospective study was conducted in 16 general hospitals of the Spanish National Health Service in 2002. All women with singleton pregnancies and without a former diagnosis of diabetes mellitus were considered for GDM screening at 24–28 weeks of gestation with a 50-g glucose challenge test (GCT). Women who had a venous plasma glucose >7.8 mmol/l were scheduled for a diagnostic, 100-g, 3-h OGTT. Both NDDG (fasting: 5.8 mmol/l, 1 h: 10.6 mmol/l, 2 h: 9.2 mmol/l, 3 h: 8.1 mmol/l) [6] and ADA criteria (fasting: 5.3 mmol/l, 1 h: 10 mmol/l, 2 h 8.6 mmol/l, 3 h: 7.8 mmol/l) were considered [5]. With both criteria, GDM was defined when more than two plasma glucose measurements were equal to or higher than the cut-off points. The term ‘ADA only gestational diabetes’ (ADA-only-GDM) was used to refer to pregnant women who would be diagnosed as having GDM by ADA but not by NDDG criteria [13].

Following the method of Magee et al. [9], four glucose tolerance groups were defined: women with GDM according to NDDG criteria receiving usual care (NDDG-GDM group) and three untreated groups, representing a gradient of carbohydrate tolerance. Negative screenees had a glucose value below 7.8 mmol/l on GCT. False-positive screenees had a positive GCT but a normal OGTT by ADA criteria and did not receive specific treatment [5]. Women with NDDG-GDM were managed with diet and self-monitoring of blood glucose. Insulin was added when required according to the local protocol (ADA recommendations [7], or lower thresholds in some centres [14]). All women received routine prenatal care.

Data collected included: maternal age, prepregnancy BMI, metabolic therapy during pregnancy, chronic hypertension (prior to pregnancy or diagnosed before 20 weeks of gestational age), PIH (when diagnosed after 20 weeks of gestational age, irrespective of proteinuria so that it also included pre-eclampsia), delivery (gestational age, spontaneous/induced, vaginal/Caesarean section), reason for Caesarean section (elective/dystocia/fetal distress/others) and newborn characteristics (birthweight, sex, Apgar score, perinatal mortality). Newborns were defined as large for gestational age (LGA) when sex-specific birthweight for gestational age was above the 90th percentile of Spanish fetal growth curves [15]. Gestational age was defined as completed weeks, based on last menstrual period or on the earliest ultrasound assessment if discordant. Before hospital discharge, newborns were examined for major congenital malformations (MCMs), defined as those causing significant functional or cosmetic impairment, requiring surgery, or being life-limiting.

Macrosomia (defined as a birthweight at or above 4 kg) and Caesarean section were defined as primary outcomes. Sample size (n=9,741) was calculated to detect a two-fold increase in the rate of macrosomia and a 40% increase in the rate of Caesarean section in the subgroup of women fulfilling ADA-only-GDM criteria, assuming baseline rates of 4 and 12% respectively (α=0.05 and β=0.80). These figures were chosen because they are the relative risks observed in the Toronto Tri-Hospital Study [4]. Secondary outcomes were the rate of LGA newborns, preterm birth, PIH, Apgar score <7 at 1 and 5 min, MCM and perinatal mortality. The projected sample size was calculated to be insufficient to detect a 70% increase in the risk of PIH in the ADA-only-GDM group (observed for pre-eclampsia in the Toronto Tri-Hospital Study) assuming a baseline rate of 2%.

Glucose tolerance groups were compared using a chi square test for qualitative variables and ANOVA for quantitative variables. When overall analyses were significant, all possible intergroup comparisons were made. Multiple logistic regression analysis with a backward method was used to calculate adjusted odds ratios (ORs) for developing complications and 95% CIs. All potentially predictive variables were fed into the model; BMI was modelled as a categorical variable and maternal age as a continuous variable. Significance was set at a two-tailed p<0.05.

Results

The study recruited 9,513 consecutive women aged 14–45 years. Ethnicity was 91.8% Caucasian, 0.7% African, 0.7% Asian, 2.5% Arab, 2.7% Caribbean and 1.6% other. We excluded 206 women (2.2%) who did not undergo screening and 37 with non-available results. The final study group was therefore made up of the remaining 9,270 (97.4%) pregnant women.

Table 1 shows maternal characteristics and plasma glucose levels in the screening and diagnostic tests. A total of 819 (8.8%) women met NDDG criteria and 263 (2.8%) also met ADA-only-GDM criteria. Applying ADA thresholds to this population, GDM prevalence increased by 31.8%. Glucose tolerance was not related to non-Caucasian ethnicity. Prevalence of chronic hypertension and mean age, weight and BMI were higher in women with worse glucose tolerance, whereas mean height was lower in these women (overall p<0.001). As expected, the rates of fetal macrosomia and Caesarean section were higher in women with lower glucose tolerance. The rate of fetal macrosomia differed between negative screenees and the groups with worse glucose tolerance (overall p<0.001), but the differences among the latter were not significant. Women with GDM according to NDDG criteria had a higher rate of Caesarean section than negative screenees (24.8 vs 19.2%, p=0.002), but this was not the case for women with ADA-only-GDM (22.5 vs 19.2%, NS). No differences in the reason for Caesarean section were seen between groups (data not shown). As to secondary outcomes, only the rates of PIH and LGA were higher in women with lower glucose tolerance (see Table 1 for details). In relation to negative screenees, preterm birth was less frequent in ADA-only-GDM and more frequent in NDDG-GDM. There were no differences among glucose tolerance groups in low Apgar score at 1 and 5 min, perinatal mortality, or MCM. Nor were differences observed among groups regarding the rate of small-for-gestational-age (SGA) newborns, a variable which was not included as an outcome, but is nevertheless related to glucose tolerance and hyperglycaemia.

Table 1 Maternal characteristics and clinical outcomes according to the glucose tolerance status
Full size table

Age and BMI were categorised in quartiles. As expected, GDM prevalence increased with age and BMI quartiles (p<0.0001 for trend in both). Each woman received a score for age and BMI corresponding to the quartile of the variable, and a combined score was constructed with the sum of the former. The lowest possible combined score was 2, corresponding to women in the 1st quartile of age and BMI, whereas the highest possible combined score was 8, corresponding to women in the 4th quartile for both age and BMI. The categories of very low, low, intermediate and high risk were defined as corresponding to a sum of 2, 3–4, 5–6 or 7–8 in this scale respectively. In the lower quartile for age (<26 years) we observed a 4.3 and 1.5% prevalence of GDM according to NDDG and ADA criteria. Except for the category with the lowest risk according to BMI and the combined score, the relative increase in GDM prevalence with ADA criteria displayed an inverse relationship with the risk category according to age (34.8–34.7–31.7–27.2%, p<0.001), BMI (33.9–41.8–32.1–25%, p<0.001) and the combined score (20–40–34.3–22.5%, p<0.001).

Significantly predictive models were constructed for the two primary outcome variables (macrosomia and Caesarean section). Glucose tolerance was a significant predictor in both models (Table 2). Nevertheless, ADA-only-GDM category (untreated) was not associated with a significantly different risk, whereas NDDG-GDM (treated) had an increased risk of macrosomia (OR 1.47, CI 1.06–2.06) and borderline risk of Caesarean section (OR 1.21, CI 0.99–1.47). Predictive models were also constructed for four secondary outcomes (PIH, preterm birth, LGA and Apgar 1′<7) with glucose tolerance being included in the four models (Table 2). Women with ADA-only-GDM had a significantly increased risk of PIH (OR 2.34, CI 1.15–4.77) and LGA newborns (OR 1.44, CI 1.02–2.03), whereas women with NDDG-GDM had a significant increase in the frequency of PIH (OR 2.03, CI 1.30–3.16) and preterm birth (OR 1.44, CI 1.04–2.00). No significant predictors were found for MCM, perinatal mortality or Apgar 5′<7.

Table 2 Multivariate analysis for prediction of primary (macrosomia, Caesarean section) and secondary (PIH, LGA, preterm birth, Apgar 1′<7) outcomes
Full size table

Discussion

This study was designed to determine the potential impact of adopting ADA criteria for GDM diagnosis on prevalence and morbidity in the Spanish population. The 8.8% prevalence of GDM according to the NDDG criteria was in the middle range of Spanish [16–20] and worldwide rates. There are unquestionable ethnic differences in GDM prevalence, with non-Caucasian women being at higher risk [8, 21–23], but in our study the high GDM prevalence was not related to non-Caucasian ethnicity. As in previous reports [21, 24, 25], higher maternal age and BMI were associated with GDM. The relatively older age of this cohort could partially explain the high GDM prevalence. However, ethnicity could be another factor. People from European Mediterranean countries define themselves as Caucasian, but in terms of diabetes mellitus prevalence, the risk differs between European Mediterranean and non-Mediterranean countries [26]. Furthermore, in their classic paper on GDM prevalence according to country of birth, Beischer et al. [27] reported the risk as being 37% higher in women from European Mediterranean countries than in those from Northern Europe.

The 11.6% overall prevalence of GDM using ADA criteria was higher than previously reported in a small Spanish study [16]. This prevalence implies a 31.8% relative increase in GDM diagnosis according to NDDG criteria. Although the increment in prevalence with ADA criteria (2.8%) was higher than that found in previous studies (0.5–1.5%), the relative increase was lower [10, 13]. Ferrara et al. [10] reported that proportional increases in prevalence were higher in subgroups at low risk according to age or ethnicity, and the present study would extend this feature to populations.

If ADA criteria were to replace NDDG criteria, the number of pregnant women diagnosed with GDM would greatly increase, thus raising the burden of prenatal care. However, the decision to use ADA thresholds should only be made to prevent maternal and perinatal complications. The Toronto Tri-Hospital Gestational Diabetes Project [4, 28] showed that women with untreated ADA-only-GDM had higher rates of Caesarean section and macrosomia than both normoglycaemic women and women with more severe NDDG-GDM who had received specific treatment. Additional studies have also justified shifting from NDDG to ADA criteria [9, 29–31], but this is not a unanimous conclusion [13].

According to multiple regression analysis, women with ADA-only-GDM differ from negative screenees in two secondary outcome variables (LGA and PIH) but no primary variables. Again, according to multiple regression analysis and vs negative screenees, women with NDDG-GDM had a higher risk of macrosomia, Caesarean section (borderline), PIH and prematurity, but not of LGA. Predictors of analysed outcome variables are consistent with current knowledge for macrosomia/LGA [32, 33], Caesarean section [34, 35], PIH [4, 36], preterm birth [37] and low Apgar score [38, 39], but we found neither increased BMI nor abnormal glucose tolerance to be predictors for MCM. With regard to Caesarean section, we should emphasise that some authors found no increased risk in ADA-only-GDM [31, 40]. The rate of PIH in the present study was similar to that found in a Spanish study published 17 years ago [13] and is within the lower range of published reports. In the negative screenees group we observed 1.7% PIH including pre-eclampsia, vs 4.9% pre-eclampsia only in the corresponding group of the Toronto Tri-Hospital Study [4], or 3.7% PIH including pre-eclampsia in the normal weight group in a study of glucose-tolerant women [41]. The rate of PIH including pre-eclampsia that we observed in the ADA-only-GDM group (3.8%) was lower than the aforementioned rate of pre-eclampsia only in the negative screenees group of the Toronto Tri-Hospital study.

We can presume that if women with ADA-only-GDM received the same metabolic therapy as those with NDDG-GDM, the risk of LGA and PIH in the former would be similar to or better than that in women with NDDG-GDM. Nevertheless, we are concerned that with tight metabolic control, the rate of SGA newborns in the ADA-only-GDM group could increase to a figure similar to or higher than that in NDDG-GDM women [42], cancelling out some of the benefits of reducing the rate of LGA newborns.

This study has two theoretical weaknesses. In contrast with Sermer et al. [4], the diagnostic strategy did not include an OGTT in every pregnant woman. We must therefore assume that we have included some women with GDM in the negative screenees group. Consequently, as we may have underestimated the overall prevalence of GDM, and particularly that of ADA-only-GDM, the relative increase in GDM prevalence with ADA criteria should be somewhat higher. In addition, as this study was not blinded, a higher risk of Caesarean section in women with ADA-only-GDM could have been attributed to a biased obstetric practice. However, this was not the case. The strength of the study is that it was adequately powered to detect a different outcome in the ADA-only-GDM group. The first column in Table 1 indicates that baseline assumptions were correct for macrosomia and PIH, but not for Caesarean section. Nevertheless, additional calculations indicate that the estimated sample size should also be sufficient to detect a 40% increase in the rate of Caesarean sections with the baseline rate observed in the study. The rate of Caesarean section was higher in women with NDDG GDM (a 5.6% increase in crude rates and borderline significance in multivariate analysis). This higher rate could be accounted for by the 2.8% increase in macrosomia and the 2.5% increase in PIH in this group. Even when this study was not specifically designed to analyse the ‘practice style effect’ [28] towards an increased rate of Caesarean section in women with a diagnosis of GDM, these figures do not support a substantial effect.

According to our data, adoption of ADA criteria would increase the prevalence of GDM by 31.8%. However, we do not consider that a change in diagnostic criteria is warranted in our setting as the contribution of these additionally diagnosed cases to adverse GDM outcomes is not substantial.

Collaborators

Additional members of the Spanish Group for the study of the impact of Carpenter and Coustan GDM thresholds were: C. Bach, Hospital Universitari de Girona Doctor Josep Trueta, Girona; M. D. Maldonado, O. Rodriguez, I. Díaz, O. Gutiérrez, A. Carazo and M. Martín, Hospital Virgen de la Salud, Toledo; M. Navarro, P. Velasco and J. Sancho–Miñano, Hospital Universitario Virgen de las Nieves, Granada; L. Blesa, M. Sánchez-Dehesa and E. Faure, Hospital Clínico Universitario Lozano Blesa, Zaragoza; M. N. Suárez, Hospital Universitario de Canarias, Tenerife; A. Santcalpe, Hospital Mutua de Terrassa; I. Rodriguez, Complejo Hospitalario Universitario Xeral-Cíes de Vigo; J. C. Martínez, S. Regel and A. Picó, Hospital General Universitario, Alicante; W. Plasencia and R. García, Hospital Universitario Materno-Infantil de Canarias, Las Palmas de Gran Canaria; T. Montoya and S. Monereo, Hospital Universitario de Getafe; P. Martín-Vaquero, L. Herranz, M. Jañez and M. J. Delgado, Hospital La Paz, Madrid; I. Mascarell, P. Cubells and R. Casañ, Hospital Universitario Dr Peset, Valencia; R. Gironés, Hospital Clínico de Valencia; and J. M. Adelantado and G. Ginovart, Hospital de Sant Pau, Barcelona.