Changing diagnostic criteria for gestational diabetes (CDC4G) in Sweden: A stepped wedge cluster randomised trial

Introduction

Gestational diabetes mellitus (GDM) is the most common medical complication of pregnancy with a growing prevalence globally, to a large extent due to the increase in overweight and obesity [1]. Hyperglycemia during pregnancy is associated with complications for mother and child during pregnancy and delivery, but also associated with raised risks of later type 2 diabetes and cardiovascular disease for the mother. For the child, there is emerging evidence about risks for future obesity and associated comorbidity [2–4]. The clinical controversy about what glycemic thresholds to diagnose and treat GDM relates to the balance between sufficient evidence of health improvements in different populations and increased workload with associated costs [5–7]. To progress towards a universal standard approach to GDM diagnosis, the World Health Organisation (WHO) adopted the International Association of the Diabetes and Pregnancy Study Group’s diagnostic criteria in 2013 [1]. These WHO-2013 criteria, with an increase in the number of women diagnosed [8], are based on risk for adverse pregnancy outcomes, in contrast to older criteria relating to maternal risk of developing type 2 diabetes postpartum. It has been unclear whether treatment using these WHO-2013 criteria improves outcomes for the mother or the child. The few randomised trials evaluating the effect of GDM treatment have been performed in different populations, with varying screening strategies and comparing different diagnostic criteria for GDM [5–7].

In Sweden, screening for GDM has mostly been performed based on clinical risk factors and repeated random plasma glucose measurements. Diagnostic criteria have been mainly those of overt diabetes (fasting ≥7.0 mmol/L, 2-h value 8.9 to ≥11.1 mmol/L) [9], resulting in a low prevalence of GDM compared with other countries [10]. In 2015, the Swedish National Board of Health and Welfare adopted the WHO-2013 criteria. A national stepped wedge cluster randomised controlled trial (SW-CRT) was considered a pragmatic approach to test whether a reduction in adverse neonatal and/or maternal outcomes could be detected following the implementation of the WHO-2013 criteria in Sweden in a real-world setting [9].

The primary aim of the Changing Diagnostic Criteria for Gestational Diabetes (CDC4G) trial was to evaluate whether implementation of the WHO-2013 criteria leads to a reduction in large for gestational age (LGA, 90th percentile) infants, and the secondary aim was to evaluate possible reduction of other adverse neonatal and maternal outcomes.

Discussion

In this SW-CRT of implementing the WHO-2013 diagnostic criteria in Sweden in a population screened using risk factors and repeated random plasma glucose measurements, there was a 2.5-fold increase in GDM diagnosis. The switch to the WHO-2013 criteria did not lead to a decrease in the primary outcome, LGA (>90th percentile) or composite neonatal or maternal outcomes in the mITT population. In the subgroup actually treated after the switch (based on fasting and 2-h values in the OGTT), no significant reduction in LGA (>90th percentile) was seen. However, there was a substantial reduction in the strength of the associations with both neonatal and maternal composite outcome; although more neonates were identified with hypoglycemia (glucose <2.6 mmol/L), without any associated increased need of intravenous glucose therapy. In the mITT population, there were adverse neonatal outcomes (respiratory distress, mechanical ventilation, and cranial hemorrhage) that are unlikely to be a result of implementation of the WHO-2013 criteria since the sample size of the subgroup discordant for definition of GDM is small (only approximately 4% of the study population) and very few or no adverse outcomes were seen in the subgroup affected by the intervention. The reduced risk for the exploratory outcomes macrosomia (≥4.5 kg) and severe LGA in both the mITT and subgroup populations are clinically important outcomes relevant for decision-making. This reduction in birth weight is likely to be related to the benefits seen in maternal composite outcome (severe hemorrhage, perineal trauma, and shoulder dystocia). A decrease in breastfeeding at discharge was noted in both the mITT and the subgroup but may be non-generalisable due to the high rate of missing values for this self-reported outcome measure.

Two previous randomised controlled trials (RCTs) have studied the change from local guidelines to the WHO-2013 GDM criteria [7,23]. The most comparable RCT to our trial, the Gestational Diabetes Mellitus Trial of Diagnostic Detection Thresholds (GEMS), was conducted in New Zealand with a two-step OGTT screening [24] and reported no reduction in the primary outcome LGA in the total obstetric population but a reduction in their subgroup [7]. Differences in growth standards, population characteristics, screening methods, former diagnostic criteria, obstetric surveillance guidelines, and treatment targets make comparisons between the trials difficult and probably explain differences in measured outcomes. In the CDC4G trial, only women with risk factors for diabetes and high BMI were tested and treated, which is one major factor probably explaining some differences in outcomes. Furthermore, in the CDC4G trial, induction of labour was performed at 40+6 weeks’ gestation at the latest for medically treated women and diet-treated women according to local guidelines up to 42+0 weeks’gestation, i.e., later than many other recommendations and guidelines [25].

Similar to the GEMS trial, we found an increased risk for neonatal hypoglycemia, likely due to surveillance bias from routine neonatal plasma glucose monitoring in neonates as previously shown [7]. However, identifying and treating more neonates with hypoglycemia might improve long-term neurocognitive outcomes [26].

The rate of preeclampsia differed between the trials within the subgroups discordant for definition of GDM, which was increased after the intervention in our trial but decreased in the GEMS trial. The increase in preeclampsia in our trial might be explained by surveillance bias but needs further evaluation. Furthermore, there were differences in breastfeeding rates at discharge between the studies: routines for supplementary feeding [27] might partly explain this difference. For example, in New Zealand, Dextrose gel was fully implemented during the period when the GEMS study was conducted [28]. In Sweden, Dextrose gel was recommended in the national guidelines for the first time in 2017 [29] and thus, not fully implemented during the CDC4G study. In addition, given the high proportion of missing values for self-reported breastfeeding in our study, these results should be treated with some caution.

Strengths include this being to the best of our knowledge, the first SW-CRT evaluating the WHO-2013 criteria enabling inclusion of approximately half of all deliveries in Sweden during 2018 in a real-world scenario with comprehensive data collection through national registries. The methodological complexities in the trial design included potential confounding with time and clusters, which were adjusted for in the analysis. The risk of selection bias is likely to be very low, as access to care is high (pregnancy care is free) and registers provided standardised medical information on all pregnancies, with coverage of >95% [15]: this makes the trial generalisable to a population screened by risk factors and also temporal trends in outcomes and/or possible residual confounding could be identified. The robustness of our data is further evident by the relatively unchanged risk after adjustments for various maternal characteristics, but residual confounding cannot be ruled out entirely. Also, the agreed treatment and surveillance guidelines that were implemented before starting, which is a major strength. We were able to implement the venous plasma sampling method across all the included clusters, and the Swedish national quality goals for glucose measurements were followed at all sites except one, including the use of fluoride citrate tubes for laboratory methods and quality assessment of patient near methods [30]. Even though 3 clusters were excluded from analysis, the study had adequate statistical power. To the best of our knowledge, this is the first major RCT of GDM criteria to use citrate to prevent ongoing glycolysis during the OGTT sampling making the glucose values more stable than using, e.g., fluoride alone [31].

Potential limitations of the trial are mainly due to the trial design. Although the planned sample size was exceeded, the power calculation was based on the assumption of the number of OGTTs generated by a universal one-step diagnostic approach [19].

We had to exclude 3 clusters (using one-step capillary OGTT screening) that were randomised, that were not able to change to venous OGTT as defined by the study protocol during the study period. This has however not introduced differential bias, since all pregnancies in all these clusters were excluded and no pregnancy from these centres was included in the mITT analysis [14]. Even though uniform treatment guidelines existed, it was impossible to control compliance with treatment and management strategies entirely. For the subgroup analysis, no comparison could be made based on the 1-h value, since the masking and introduction of a 1-h value in the OGTT was not possible to implement. As the duration of the study was only 1 year, we were unable to fully account for the seasonal variation which might include LGA and glucose values [32,33]. There was also a risk of chance positive findings due to multiple testing among the prespecified outcomes.

Concerns about implementing the WHO-2013 criteria have been raised previously [34], including increased resources and costs. The economic consequences have been analysed in conjunction with the CDC4G trial and will be reported separately. Whether the introduction and treatment of the WHO-2013 criteria result in long-term health benefits for mother or child, needs to be evaluated in future follow-up studies in different populations [2–4,35].

Diagnosis of GDM increases the likelihood that women will attend postpartum follow-up programmes and may help to prevent future type 2 diabetes and cardiovascular disease.

The findings of the CDC4G trial must be placed within the wider discoveries in GDM research over the last 2 to 3 years. Like previous trials [7,36,37], the CDC4G trial found no benefit in reducing LGA defined as birth weight >90th percentile on a population level [16], but we could show a reduced risk for macrosomia ≥4.5 kg and LGA +2 SD [16] in the total pregnant population that most likely leads to the reduction in perineal trauma and shoulder dystocia. As with the GEMS study [7], a larger benefit occurred within the subgroup of women treated based on WHO-2013 criteria. This raises the question of why we treat GDM: for the benefit of the at-risk mother and baby or for the total obstetric population? Also, risk factor and random plasma glucose screening, with its lower sensitivity, denies many women GDM treatment [38,39] and the possibility to avoid adverse pregnancy outcomes. Beyond this, to the best of our knowledge, the CDC4G trial was the first to use citrate in a large trial for GDM, suggesting that treatment was actually commenced at a threshold below the HAPO cutoffs [31]. The implications of using more stable glucose sample handling, and other changes beyond thresholds, also need to be considered in any future changes in approach to diagnosing GDM. Finally, the recent findings of the TOBOGM study [40] suggest that new approaches to GDM screening and treatment need to include early detection and higher thresholds than the WHO-2013 criteria, since a raised risk for SGA was reported with treatment of lower levels of hyperglycemia (those treated between cutoffs for OR 1.75 and OR 2.0 from the HAPO study) [41] in early pregnancy in the TOBOGM study. With this new knowledge, it is obvious that the current screening and diagnostic approaches need to be reviewed: this represents an excellent opportunity for a coherent approach. We hope that this study, together with the other major RCTs and new scientific evidence, will contribute to the process that the Swedish National Board of Health and Welfare and other professional bodies in Sweden are working on to make changes in both screening, definition, and treatment of hyperglycemia during pregnancy. To date, to the best of our knowledge, no RCT has evaluated pregnancy outcomes after the implementation of the WHO-2013 criteria in a setting where universal one-step 75 g OGTT screening has been used. New technology and possible biological markers might be helpful in simplifying screening procedures and working towards a more individual approach in both identification and treatment of hyperglycemia during pregnancy for prevention of adverse outcomes in both the short and long term for mother and child.

Implementing the WHO-2013 diagnostic criteria for GDM in a risk factor-based screening setting did not reduce the risk of the primary outcome LGA (>90th percentile) in the total population or the subgroup affected by treatment. However, there was an associated reduction in adverse neonatal and maternal outcomes, with the largest effect in the subgroup of women whose OGTT results were discordant between the old and new criteria for the definition of GDM.