The Belgian Diabetes in Pregnancy Study (BEDIP-N) was a multicentre prospective cohort study that has previously been described in detail [6,7,8]. The study protocol was approved by the institutional review boards of all participating centres. Participants provided informed consent before inclusion in the study. Women were enrolled at between 6 and 14 weeks of pregnancy. Those without impaired glucose tolerance or diabetes in early pregnancy (defined by ADA criteria) received both a non-fasting 50 g glucose challenge test (GCT) and a 75 g 2 h OGTT between 24 and 28 weeks of pregnancy. During the study, participants and healthcare providers were blinded to the result of the GCT. Therefore, all participants received an OGTT irrespective of the GCT result. The diagnosis of GDM was based on the International Association of Diabetes and Pregnancy Study Groups criteria, now commonly referred to as the 2013 WHO criteria for GDM [6, 7]. Of all the BEDIP-N participants, 1813 (90.3%) received both a GCT and an OGTT between 24 and 28 weeks of pregnancy, with 228 women diagnosed with GDM (prevalence of 12.4%). Overall, 106 participants (5.3%) discontinued the study before 24 weeks of pregnancy, half because of a medical reason [8]. We have recently shown that the threshold of the GCT would need to be reduced to at least 7.2 mmol/l to achieve sensitivity of ≥70% for GDM based on the 2013 WHO criteria [7]. Based on a GCT result of ≥7.2 mmol/l and the 75 g OGTT, women without GDM could be stratified into the following two groups: NGT on the antepartum OGTT with an abnormal preceding GCT result (abnormal GCT NGT group); and NGT on the OGTT with a normal preceding GCT result (normal GCT NGT group).
The ADA-recommended glycaemic targets were used for the treatment of GDM [9]. If targets were not met within 2 weeks after the start of lifestyle measures then insulin was started. Women with GDM were invited for an extra visit at 6–16 weeks postpartum to receive a 75 g OGTT, and the ADA criteria were used to define diabetes and glucose intolerance (impaired fasting glucose and/or impaired glucose tolerance) [6, 9].
Baseline characteristics and the obstetric history were collected in early pregnancy [6]. An ethnic minority background was defined as having at least one parent from a non-white origin. In early pregnancy and at 24–28 weeks of pregnancy, anthropometric measurements were obtained and several self-administered questionnaires were completed [6]. BP was measured twice at 5 min intervals using an automatic BP monitor [6]. A woman was defined as being overweight if her BMI was ≥25 kg/m2 and as obese at BMI ≥30 kg/m2.
At the first visit between 6 and 14 weeks of pregnancy, a fasting blood test was performed to measure fasting plasma glucose (FPG), insulin, lipid profile (total cholesterol, HDL- and LDL-cholesterol, triacylglycerols) and HbA1c. HOMA-IR and HOMA-B were measured in early pregnancy, as previously described [10]. At the time of the OGTT, the fasting lipid profile and HbA1c were measured. Glucose and insulin were measured fasting and at 30, 60 and 120 min. An increase in triacylglycerols was defined as the difference between fasting triacylglycerol levels in early pregnancy and at the time of the OGTT [6].
To define GDM subgroups based on different degrees of insulin resistance, we used insulin and glucose levels during the 75 g OGTT at 24–28 weeks of pregnancy to calculate the Matsuda index, a whole-body measure of insulin sensitivity [11]. For the main analyses, we classified GDM in two different groups based on the Matsuda index: a GDM insulin-resistant group, defined as a Matsuda index <50th percentile (P) of the NGT group; and a GDM insulin-sensitive group, defined as a Matsuda index above P50 of the NGT group. In the electronic supplementary material (ESM), we provide additional analyses of four GDM subgroups stratified according to the Matsuda index of <P25, P25–P50, P50–P75 and >P75 compared with women with NGT (ESM Tables 1–3). In addition, HOMA-IR and indices of beta cell function (HOMA-B, the insulinogenic index divided by HOMA-IR, the insulin secretion-sensitivity index-2 [ISSI-2] and the Stumvoll index) were measured, as previously described [10, 12,13,14].
The following pregnancy outcome data were collected: gestational age, pre-eclampsia (de novo BP ≥140/90 mmHg for >20 weeks with proteinuria or signs of end-organ dysfunction), gestational hypertension (de novo BP ≥140/90 mmHg for >20 weeks), type of labour and type of delivery with the indications, birthweight, macrosomia (>4 kg), birthweight ≥4.5 kg, large for gestational age (LGA) infant (defined as birthweight >P90 according to standardised Flemish birth charts adjusted for the sex of the baby and parity [15]), small for gestational age infant (defined as birthweight <P10 according to standardised Flemish birth charts adjusted for the sex of the baby and parity [15]), preterm delivery (<37 completed weeks), 10 min Apgar score, shoulder dystocia, neonatal respiratory distress syndrome, neonatal jaundice, congenital anomalies and admission to the neonatal intensive care unit (NICU) [6]. A glycaemic value of <2.2 mmol/l, irrespective of the need for i.v. administration of glucose and admission to the NICU, was considered as a neonatal hypoglycaemia across all centres. Admission to the NICU was decided by the neonatologist in line with the normal routine of each centre. Excessive weight gain during pregnancy was defined according to the 2009 Institute of Medicine guidelines [16]. Early weight gain was calculated as the difference in weight between the first prenatal visit and the time of the OGTT, while total weight gain was calculated as the difference in weight between the first prenatal visit and delivery.
Analyses of FPG at 6–14 weeks and the glucose measurements of the OGTT were conducted locally at each centre. Analyses of GCTs and insulin, lipid and HbA1c levels were conducted centrally at the laboratory of UZ Leuven, and these results were not communicated to participants or healthcare providers during the study. Plasma glucose was measured using an automated colorimetric–enzymatic method on a Hitachi/Roche Modular P analyser (Basel, Switzerland). Insulin was measured using an immunometric electrochemiluminescence immunoassay analyser (Roche Modular E170, Roche, Basel, Switzerland). HbA1c was measured using a Tosoh Automated Glycohemoglobin Analyzer HLC-723G8 (Tokyo, Japan). Lipid levels were measured using the immunoassay analyser Cobas 8000 (Roche). Coefficients of variance were 1% for glucose, 6% for insulin, about 2% for lipids and 2% for HbA1c in the laboratory of UZ Leuven.
Statistical analysis
Characteristics and pregnancy outcomes were compared between each GDM subtype and the normal GCT NGT group. The frequency of glucose intolerance postpartum was evaluated in women with GDM. Continuous variables are presented as means ± SD if normally distributed and as medians (interquartile range) otherwise, and categorical variables as percentages. The χ2 test (or Fisher’s exact test in case of small cell frequencies) was used for comparing categorical variables between groups. The Mann–Whitney U test or Kruskal–Wallis test was used to compare continuous variables among two or multiple groups, respectively. Additional comparisons of pregnancy outcomes between the GDM insulin-resistant group (and the GDM high insulin resistance group in the additional analyses in ESM Table 3) and the NGT group were performed using multivariable logistic regression analyses adjusted for the following confounders: model 1: unadjusted model; model 2: adjusted for demographics (centre, age, ethnicity, parity, education and smoking before pregnancy); model 3: model 2 plus adjusted for first-degree family history of diabetes, previous history of GDM and BP in early pregnancy; model 4: model 3 plus adjusted for maternal BMI and waist circumference in early pregnancy; model 5: model 4 plus adjusted for fasting lipids (total cholesterol, LDL-cholesterol and triacylglycerols), FPG and HbA1c in early pregnancy; model 6: model 5 plus adjusted for gestational weight gain until the OGTT at 24–28 weeks of pregnancy.
A p value of <0.05 (two-tailed) was considered significant. Given the large number of statistical tests performed, no correction for multiplicity was applied in order to maintain sufficient power. Analyses were performed by A. Laenen using SAS software (Cary, NC, USA, version 9.4).