Improved pregnancy outcomes in women with type 1 and type 2 diabetes but substantial clinic-to-clinic variations: a prospective nationwide study

Pregnancy in women with diabetes is associated with increased risks of serious adverse outcomes with a two-to-five fold increased risk of congenital anomaly, stillbirth and neonatal death compared with the general maternity population [1‐3]. Less severe but more frequent perinatal complications relating to maternal diabetes include preterm delivery, large for gestational age (LGA) infants and neonatal intensive care unit admission, affecting approximately one in two babies [4, 5]. Nationwide studies from the Netherlands, Sweden and Finland suggest no improvement in either rates of serious adverse pregnancy outcomes or perinatal complications in recent decades, with possible explanations including increasing maternal age, longer duration of diabetes and increasing rates of overweight and obesity among women of reproductive years.

A 2002/2003 population-based study, conducted by the Confidential Enquiry into Maternal and Child Health (CEMACH), examined the quality of maternity care and documented pregnancy outcomes in the UK among women with diabetes, concluding that pregnancy preparation was inadequate, resulting in potentially modifiable poor pregnancy outcomes in both type 1 and type 2 diabetes [6]. Since then the National Institute for Health and Care Excellence (NICE) have developed diabetes pregnancy guidelines with a clear emphasis on improving provision of prepregnancy and antenatal diabetes care [7]. The NICE guideline recommendations for prepregnancy preparation include taking 5 mg preconception folic acid, presenting for antenatal care before 8 weeks’ gestation and avoiding potentially harmful medications. The 2015 update lowered the maternal glycaemic control target from HbA_1c < 7.0% (53 mmol/mol) to HbA_1c < 6.5% (48 mmol/mol) and recommended elective delivery between 37⁺⁰ and 38⁺⁶ weeks’ gestation [8].

Responding to the NICE guidelines, a National Pregnancy in Diabetes (NPID) audit was established to document the pregnancy preparation, antenatal care and fetal health outcomes for pregnant women with type 1 and type 2 diabetes [9].

Here we present pregnancy outcomes of women with diabetes who delivered between 1 January 2015 and 31 December 2015. Our aims were to provide contemporary data on the rates of serious adverse pregnancy outcomes (congenital anomaly, stillbirth and neonatal death) and perinatal complications (preterm delivery, LGA and neonatal care admission), and to explore the relationship between maternal deprivation and clinic-to-clinic variations with maternal glycaemic control and folic acid supplementation. We also describe changes since the previous 2002/2003 CEMACH survey.

Healthcare professionals at each maternity unit completed standardised web-based data entry forms for every pregnant woman with diabetes who delivered between 1 January and 31 December 2015. As not all pregnancies resulted in a delivery these numbers included pregnancies ending in miscarriage or termination within this date range. All maternity units in England, Wales and the Isle of Man providing antenatal diabetes care were expected to participate (http://​digital.​nhs.​uk/​npid). Data were obtained only from women who provided written informed consent. The information leaflet and consent form met the Health Research Authority requirements for clinical audit, and research ethics approval was not required. Linkage with data collected in other systems (Hospital Episodes Statistics data, Patient Episode Database for Wales, National Diabetes Audit) allowed only a limited number of data items to be collected by local teams. This included information regarding type of diabetes, medications and folic acid use before conception. Data regarding pregnancy outcome was completed locally at 28 days after delivery, miscarriage or termination.

We defined pregestational diabetes as diabetes that had been diagnosed before pregnancy, and excluded women with diabetes who presented during pregnancy. Women with monogenic diabetes, or where there was doubt about whether they had type 1 or type 2 diabetes, or the type of diabetes data item was missing were classified as ‘other’ (these pregnancies were excluded from analyses comparing type 1 and type 2 diabetes). Glycaemic control was derived from HbA_1c measurements, and the first and last recorded values during pregnancy were collected. HbA_1c was measured locally using DCCT-aligned assays. We explored the relationship between maternal deprivation with glycaemic control and folic acid supplementation using an index of multiple deprivation score for women living in England and Wales where postcode details were recorded in the National Diabetes Audit [10]. For data protection of potentially sensitive information in the clinic-to-clinic comparisons, we included only clinics with at least ten completed pregnancy records; therefore, 130 clinics were included for type 1 diabetes and 103 clinics for type 2 diabetes comparisons.

We defined stillbirth as a fetal loss occurring after 24 weeks’ gestation, and neonatal death as death of a live born infant up to 28 days after delivery. We collected data on congenital anomalies for live births and terminations of pregnancy at any gestation, including for stillbirths and for fetal loss after 20 weeks’ gestation. The reported diagnoses for congenital anomaly were obtained from the hospital ICD-10 codes (www.​who.​int/​classifications/​icd/​en/​). We calculated the congenital anomaly rate as the number of infants with one or more congenital anomalies divided by the number of live births, terminations, stillbirths and fetal loss after 20 weeks’ gestation. Infant birthweight was adjusted for maternal ethnicity, height and weight, infant sex, and gestational age at delivery for singleton pregnancies using customised centiles with large and extreme LGA defined as >90th and >97.7th percentiles, respectively (GROW Centile Calculator v5.7.7.1, Gestation Network, www.​gestation.​net) [11].

Variables which were not normally distributed are given as median (interquartile range [IQR]), while normally distributed variables are given as mean (SD). Univariate analyses comparing the proportions between groups were performed using z tests and t tests for comparing continuous variables. We used Stata 8.0 for analyses and Poisson distribution to obtain 95% CIs for the rate and prevalence ratios.

This large national study details the pregnancy outcomes of women with diabetes more than a decade after the 2002/2003 CEMACH survey. While the pregnancy outcomes of women with diabetes were still suboptimal compared with the general maternity population, substantial progress has been made with significant reductions in stillbirths in both type 1 and type 2 diabetes. There were fewer preterm deliveries, LGA infants and neonatal care unit admissions in women with type 2 compared with type 1 diabetes, reflecting the better antenatal glycaemic control in type 2 diabetes. Glucose control in women with type 1 diabetes was suboptimal, with persistently high rates of preterm delivery, LGA and neonatal care unit admissions. The highest rates of neonatal complications were seen in women with HbA_1c levels ≥6.5% (48 mmol/mol) after 24 weeks’ gestation.

Over the past decade the proportion of pregnant women with type 2 diabetes has risen from 28% to 46% of all diabetes pregnancies, with type 2 now outnumbering type 1 diabetes pregnancies in some metropolitan areas. The proportion of women taking preconception folic acid supplementation has increased both in type 1 and type 2 diabetes, which is in contrast to some recent data suggesting a small decline (from 35% to 31%) in the general maternity population [13].

Our data demonstrate the ongoing healthcare inequalities between women with type 1 and type 2 diabetes and the striking contribution of maternal deprivation in type 1 diabetes. In women with type 1 diabetes, 75% of those living in the most socioeconomically advantaged areas took 5 mg preconception folic acid and 25% achieved target HbA_1c levels in early pregnancy, whereas only 37% of women living in the most disadvantaged areas took folic acid and only 10% achieved target HbA_1c levels in early pregnancy. This information highlights the need to more effectively target prepregnancy care to women living in deprived areas. Previous studies have shown that targeting all women of reproductive years can minimise the impact of maternal deprivation on prepregnancy care attendance, folic acid supplementation and early pregnancy glycaemic control [14].

It is disappointing that so few women with type 1 diabetes achieved the NICE recommended glycaemic control target for HbA_1c in early and late pregnancy (16.1% and 40.0% respectively). We speculate that this is due to previously described physiological and pharmacological challenges of matching pre-meal insulin boluses to postprandial glucose profiles in early and late gestation [15, 16]. We have shown that while women achieving target glycaemic control are older, leaner and more socially advantaged, there is substantial variation across clinics, suggesting that the impact of clinic context is also important. Further work is needed to understand whether these clinic-to-clinic variations relate to clinic size and/or staff experience of implementing newer technologies such as insulin pump therapy and continuous glucose monitoring before and during pregnancy.

While there is much emphasis on improving preconception and first trimester glucose control to reduce congenital anomaly, stillbirth and neonatal death, our data also indicate the importance of optimal glucose control in late gestation for reducing the rates of preterm delivery, LGA infants and neonatal intensive care unit admission. An important take home message for women with unplanned pregnancy and/or suboptimal early glucose control is that achieving HbA_1c < 6.5% (48 mmol/mol) after 24 weeks was very effective for reducing perinatal complications both in type 1 and type 2 diabetes.

The absolute risk of stillbirth has consistently remained 10–13/1000 over three diabetes pregnancy audit years, which now includes 6675 diabetes pregnancies [17]. Recent data from Sweden have comparable absolute stillbirth rates (15/1000) among women with type 1 diabetes [1]. However, the Swedish diabetes rates are still five times higher than the low background maternity population rates (3/1000) characteristic of high-income countries. In 2002/2003 there was a similar fivefold increase in stillbirths among women with diabetes in the UK [3].

The reductions in stillbirths in type 1 and type 2 diabetes are not solely due to improvements in the general maternity population [12], as the UK currently reports one of the slowest rates of decline in stillbirth of approximately 17% (from 5.3 to 4.4/1000 births) over the past decade [18]. Although there are clear maternal risk factors, including previous stillbirth, multiple pregnancy, nulliparity, diabetes, maternal age over 40 years, non-White ethnicity, smoking and obesity, most of the variability in stillbirth rates is independent of established risk factors [19, 20]. We can only speculate as to the possible explanations for recent improvements, which may include earlier elective birth recommendations (37⁺⁰ weeks’ gestation), tighter glycaemic control targets (HbA_1c < 6.5% or 48 mmol/mol) and/or greater concentrations of women with diabetes among fewer maternity clinics.

Stillbirth is the most common cause of perinatal death [21], so it is surprising that the reduction in stillbirths was not accompanied by a reduction in the neonatal death rate (8.1 vs 9.6/1000 for type 1 and 11.4 vs 9.5/1000 for type 2) since 2002/2003. The reasons for this are unknown and will require further evaluation in larger datasets over longer time frames.

Our study is one of the largest to describe the pregnancy outcomes associated with contemporary diabetes care and includes over 3000 women with pregnancies complicated by diabetes. Our large sample size provides the statistical power for evaluating trends in serious but rare outcomes. Other large-scale studies conducted over 10–20 years in Sweden [1] and Finland [5] are less relevant to current clinical practice, and in particular to the increasing problem of type 2 diabetes in pregnancy [22]. Our cohort is larger than any previous studies of pregnant women with type 2 diabetes and confirms meta-analyses findings that there are no differences in the rates of congenital anomalies, stillbirths and perinatal deaths between type 1 and type 2 diabetes [23]. We do confirm important differences in the perinatal outcomes of type 1 and type 2 diabetes pregnancy with significantly lower rates of preterm delivery, LGA infants and neonatal intensive care unit admissions in type 2 diabetes.

Our study should be interpreted in the context of potential limitations. First, these are cross-sectional analyses, which preclude us from making causal inferences. Second, use of routinely collected data means we have little control over errors during data collection and variations due to differences in timing and laboratory methods for HbA_1c levels. Third, because of the information governance requirements and the pre-specified nature of these analyses, we are unable to analyse whether there are significant, independent effects of HbA_1c in early and late pregnancy and/or maternal BMI on neonatal outcomes. Furthermore, we cannot guarantee that data from all women and all maternity clinics are included, and we lack information for women who did not provide consent. The numbers of women who refuse consent is anecdotally very small. For 94 clinics that responded to an electronic survey, an average of one woman per clinic was unable to and/or refused consent (C. Cartwright, P. Curnow, C. Sylvester, B. Young, unpublished data). Having 3044 pregnancies in women providing consent is at least comparable to the 2359 pregnancies without consent in 2002/2003. We also cannot be certain that the 87 women where the type of diabetes is classified as ‘unknown’ did not have type 1 or type 2 diabetes.

More work is needed to ensure that women with type 2 diabetes and their community healthcare providers are aware of the importance of safe effective contraception and/or prepregnancy care. Further research is needed to better understand the impact of clinic size and whether centralisation in fewer, larger clinics would improve glucose control and pregnancy outcomes. It remains to be seen whether recent advances in continuous glucose monitoring and closed-loop insulin delivery will be effective for improving late gestation glucose control and reducing perinatal morbidity in type 1 diabetes pregnancy [24, 25]. For stillbirth, research to identify better predictors of placental dysfunction such as erythropoietin, pregnancy associated plasma protein A, alpha fetoprotein and angiogenic/antiangiogenic factors in women with diabetes is needed [26‐28].

The NPID data provides information that can be used at the level of individual maternity clinics so service users, healthcare professionals and funders can make informed healthcare choices. It has demonstrated substantial nationwide progress in diabetes stillbirth reductions but highlights ongoing challenges to consistently improve glucose control and reduce perinatal complications in type 1 diabetes pregnancy.