figure b

Introduction

Preterm delivery is the most important contributor to perinatal morbidity and mortality [1] and is associated with long-term impairment in the infants, e.g. learning- and motor disabilities lasting into adulthood [1, 2]. Preterm delivery occurs in up to 40% of women with type 1 and type 2 (pre-existing) diabetes, which is four times more prevalent than in the background population [3,4,5,6].

In women with type 1 diabetes, most preterm deliveries are on obstetric indications [7, 8], while spontaneous onset of preterm labour is most common in healthy pregnant women [7].

In women with pre-existing diabetes, poor glycaemic control [9,10,11], pre-existing kidney involvement [12], development of preeclampsia [8, 13,14,15] and excessive gestational weight gain [16] have previously been associated with preterm delivery, but it is not clear whether these variables are independent of each other. During antenatal care, glycaemic control and gestational weight gain can be optimised by diet, physical activity and medical treatment, while the risk of preeclampsia may be modified with strict glycaemic control, prophylactic aspirin [17] and/or antihypertensive treatment [13].

The direct indications for preterm delivery may include fetal overgrowth, suspected fetal hypoxia, preeclampsia, fetal growth restriction and oligohydramnios, but causes of preterm delivery among women with pre-existing diabetes are sparsely described in the literature [7, 8]. Falling insulin requirement in late pregnancy may reflect placental insufficiency in women with type 1 diabetes [15], but the clinical importance of falling insulin requirement leading to preterm delivery is not well studied.

The aim of this study was to identify potentially modifiable risk factors and direct causes of preterm delivery in women with pre-existing diabetes.

Methods

Study population

This study was a secondary analysis of a prospective observational cohort study on hypertensive disorders evaluated with home BP in pregnant women with pre-existing diabetes [18]. The cohort was restricted to consecutive singleton pregnancies in women with pre-existing diabetes who followed ante- and perinatal care at the Center for Pregnant Women with Diabetes, Rigshospitalet, Denmark from September 2015 to February 2018, originating from a geographically well-defined region of approximately 2.7 million inhabitants. The main exclusion criteria were age <18 years, insufficient Danish language skills and a recurrent pregnancy during the study period. The cohort has previously been described in detail [18]. Out of 296 women eligible for inclusion, 244 women (82%) accepted inclusion, and 41 women were excluded due to early fetal loss (n = 18), withdrawal of consent (n = 22) or moving to another part of Denmark (n = 1), leaving a cohort of 203 women for analysis.

Data collection and definitions

Preterm delivery was defined as delivery before 37 completed gestational weeks (259 days), with gestational age based on the ultrasound measured fetal crown-rump length at 12 gestational weeks.

Possible predictors of preterm delivery were prospectively collected from the records and included maternal baseline characteristics, HbA1c in early and late pregnancy, ultrasound estimated fetal growth and development of hypertensive disorders such as gestational hypertension and preeclampsia.

To evaluate the causes of preterm delivery, birth records of women delivering preterm were reviewed independently by two of the authors (JCS and PD). The causes of preterm delivery were grouped as ‘indicated preterm delivery based on the obstetrician’s judgement’ (indicated preterm delivery) or ‘spontaneous preterm delivery’, sub-divided into ‘premature pre-labour rupture of membranes’ or ‘onset of contractions’. Indications for indicated preterm delivery were often multifactorial and categorised into four subtypes according to one of the following most prominent factors: suspected fetal asphyxia, hypertensive disorders, fetal overgrowth and maternal features such as psychological factors or poor glycaemic control in late pregnancy. Suspected fetal asphyxia included maternal perception of decreased fetal movements, falling insulin requirement, abnormal non-stress testing (cardiotocography) and abnormal fetal haemodynamics, defined as abnormal flow in the umbilical cord and/or middle cerebral arteries.

At the first antenatal visit, maternal data on ethnicity, height, pre-pregnancy weight, smoking status, educational level, medication, parity, HbA1c, office BP, presence of hypertension and kidney involvement were registered. Pre-existing hypertension was defined as known pre-pregnancy hypertension or hypertension detected before 20 gestational weeks with an office BP ≥135 mmHg systolic and/or ≥85 mmHg diastolic and a home BP ≥130 mmHg systolic and/or ≥80 mmHg diastolic, when home measurements were available [18, 19]. Pre-existing kidney involvement was defined as the presence of microalbuminuria or diabetic nephropathy before or at the first antenatal visit. Microalbuminuria was defined as albumin/creatinine ratio (ACR) of 30–300 mg/g and diabetic nephropathy as ACR >300 mg/g [20] based on at least two values within these limits. Diabetic retinopathy was assessed by retinal photo screening.

Data on fetal head circumference, abdominal circumference and femur length were collected from ultrasound scans at 27 and 33 gestational weeks. The estimate of fetal weight was recorded from the ultrasound patient record system Astraia based on Hadlock’s algorithm [21] using measurement of head circumference, abdominal circumference and femur length. Estimated fetal weight deviation was defined as percentage deviation from the mean of estimated fetal weight of a healthy population adjusted for gestational age. Fetal overgrowth was defined as estimated fetal weight deviation exceeding +22% corresponding to +2 SD. Excessive abdominal circumference was defined as an abdominal circumference exceeding +2 SD.

Development of gestational hypertension and preeclampsia was recorded. Preeclampsia was defined as hypertension developed after 20 gestational weeks with coexistence of proteinuria, defined as ≥ +1 on a sterile urine dipstick, or symptoms from other organs [22]. Gestational hypertension was defined as hypertension diagnosed after 20 gestational weeks without fulfilling the criteria for preeclampsia [18]. Hypertensive disorders in pregnancy were defined as at least one of the following hypertensive conditions: pre-existing hypertension, gestational hypertension or preeclampsia [18].

At the last antenatal visit, gestational weight gain, HbA1c and office BP were registered. Data on maternal weight gain per week were defined as the last registered weight during pregnancy minus the weight before pregnancy divided by the total number of gestational weeks [23]. Excessive gestational weight gain was defined as exceedance of the recommended gestational weight gain according to pre-pregnancy BMI recommended by the US Institute of Medicine [24, 25].

Poor glycaemic control in late pregnancy that could indicate preterm delivery was based on individual judgement and included HbA1c, fluctuating glucose values and clinically problematic hypoglycaemia. Falling insulin requirement was defined as a decline in daily insulin dose of more than 20% from maximal dose in late pregnancy.

Gestational age at delivery (weeks + days), delivery mode, birthweight and sex of the offspring were recorded. Birthweight SD score was used to describe how far offspring birthweight was from the mean of the Nordic background population adjusted for gestational age and sex [26]. Large and small for gestational age (LGA and SGA) infants were defined as birthweight >90th and <10th percentile adjusted for gestational age and sex, respectively [26].

The following variables reflecting neonatal morbidity were recorded: Apgar score, arterial umbilical cord pH, admission to neonatal care unit and plasma glucose of the newborn within 2 h of delivery. A plasma glucose value in the offspring below 2.2 mmol/l denoted hypoglycaemia.

Routine diabetes and pregnancy care

All women followed the routine antenatal diabetes care programme for pregnant women with pre-existing diabetes focusing on strict glycaemic control, appropriate gestational weight gain and strict BP control according to Danish national guidelines [13, 18, 27, 28].

At the first antenatal visit, all women were recommended a low glycaemic index diet with a moderate intake of carbohydrates. All women were instructed in the use of carbohydrate counting with a recommended daily intake of 175 g of carbohydrates [29].

Throughout pregnancy the women were recommended to self-monitor plasma glucose before and 90 min after each main meal and at bedtime to obtain a plasma glucose value of 4.0–6.0 mmol/l preprandially and 4.0–8.0 mmol/l postprandially. The HbA1c targets were <50 mmol/mol (6.7%) before 20 gestational weeks and <40 mmol/mol (5.8%) after 20 gestational weeks [18].

In women with type 1 diabetes, mealtime insulin treatment based on carbohydrate intake using carbohydrate-to-insulin ratio and insulin correction factors were frequently used.

In women with type 2 diabetes, oral hypoglycaemic agents were discontinued at the first antenatal visit and substituted with basal-bolus insulin if indicated, according to our national guideline [28].

Approximately every 2 weeks the women consulted a diabetes specialist where insulin doses were titrated. The women were taught self-adjustment of carbohydrate-to-insulin ratio and insulin dose between the visits.

Appropriate gestational weight gain was recommended according to pre-pregnancy BMI as follows: women with a BMI <25 kg/m2 were recommended to gain 10–15 kg, if BMI was 25–29.9 kg/m2 the recommendation was 5–8 kg and if BMI was ≥30 kg/m2 the recommendation was 0–5 kg [23, 30].

According to local guidelines, during the study period prophylactic aspirin (75–150 mg/day) was recommended daily from 10 to 36 gestational weeks in women with the following risks of developing preeclampsia: preeclampsia in a previous pregnancy, diabetic kidney involvement or pre-existing hypertension [18].

All women were carefully instructed to measure home BP for 3 days with three measurements in the morning and in the afternoon (18 measurements in total), in early pregnancy and at 36 gestational weeks. In addition, home BP measurements for 3 days were recommended on an individual basis when office BP was elevated (systolic BP ≥135 and/or diastolic BP ≥85 mmHg) on review with healthcare professionals. The BP device Microlife BP A3 Plus (Microlife, Switzerland) was used [18].

The women attended obstetric visits at 8, 12, 20, 27, 33 and 36 gestational weeks including ultrasound examinations and documentation of glycaemic control, office BP and weight, and a sterile urine was screened for proteinuria with a dipstick. Fetal growth was routinely assessed at 27, 33 and 36 gestational weeks with ultrasound and more frequently if growth deviations were observed.

Maternal assessment of daily fetal activity was specifically recommended from 28 gestational weeks, and the women were advised to contact the labour ward immediately in case of decreased fetal activity for further evaluation by non-stress tests and/or ultrasound-based flow measurements [31]. Weekly routine non-stress tests were performed from 33–35 gestational weeks. Intrauterine growth restriction, fetal overgrowth, hypertensive disorders or falling insulin requirement in late pregnancy led to surveillance of fetal wellbeing by non-stress testing and/or ultrasound examinations evaluating fetal haemodynamic and fetal growth.

The study was conducted in accordance with the Helsinki declaration and approved by the local ethical committee of the capital region of Denmark (H-15019186) and the local data agency (2012-58-0004, RH-2015-289, I-Suite: 04305). Written informed consent was obtained from all participants.

Statistical analysis

Continuous data were given as mean (± SD) or median (range) unless otherwise stated, while categorical variables were given as numbers (%). Clinical data were available for ≥95% of the women unless otherwise stated in the tables. Multiple imputation of missing data was not used.

Women with and without preterm delivery were compared in the total population and for type 1 and type 2 diabetes separately. Continuous skewed data were log-transformed after which all variables satisfied normality. Independent t test was used for continuous variables, and χ2 test or Fischer’s exact test was used for categorical variables. Univariate logistic regression analysis was conducted with preterm delivery as a dependent variable. To identify independent risk factors for preterm delivery, multivariate logistic regression analysis was applied including independent variables with a priori known clinical significance or based on a p value <0.10 in the univariate analysis. Based on the number of women included in the smallest group (n = 55 with preterm delivery), the number of variables was restricted to the following six: type 1 diabetes (yes/no), HbA1c in late pregnancy (mmol/mol), kidney involvement in early pregnancy (yes/no), estimated fetal weight deviation at 27 gestational weeks (%), gestational weight gain (100 g/week) and preeclampsia (yes/no). Results are given as OR and 95% CI.

The statistical analyses were made using IBM SPSS statistics 25 (SPSS, Chicago, IL, USA). A two-sided p value <0.05 was considered statistically significant.

Results

In total, 27% (n = 55) of the 203 included women with pre-existing diabetes delivered preterm (Table 1) at median 36 + 0 weeks (Table 2). Ten of the women (5%) delivered before 34 gestational weeks.

Table 1 Maternal characteristics according to preterm (<37 weeks) or term delivery in 203 women with pre-existing diabetes
Full size table
Table 2 Fetal growth and pregnancy outcome according to preterm (<37 weeks) or term delivery in 203 women with pre-existing diabetes
Full size table

Women delivering preterm had a higher prevalence of pre-existing kidney involvement and pre-existing hypertension (Table 1). HbA1c in early pregnancy was 51 ± 10 mmol/mol (6.8%) in women delivering preterm vs 49 ± 10 mmol/mol (6.6%) in women delivering at term (p = 0.22), and in late pregnancy 44 ± 7 (6.2%) vs 42 ± 6 mmol/mol (6.0%) (p = 0.055) (Table 1). All except nine (96%) women had HbA1c <53 mmol/mol (7.0%) in late pregnancy.

Preeclampsia was present in 26% of women delivering preterm and in 5% of the remaining women (p < 0.001), while the prevalence of gestational hypertension was comparable between the two groups. Among women delivering preterm there was a higher positive fetal weight deviation at 27 gestational weeks and seven (13%) presented with excessive fetal abdominal circumference (Table 2). Weekly gestational weight gain was on average 21% higher in women delivering preterm compared with women delivering at term (p = 0.03) (Table 1).

Univariate logistic regression analysis identified type 1 diabetes, duration of diabetes, higher positive ultrasound estimated fetal weight deviation at 27 gestational weeks, fetal abdominal circumference >+2 SD at 27 gestational weeks, higher gestational weight gain, kidney involvement and preeclampsia as risk factors for preterm delivery, while neither HbA1c in early nor late pregnancy were risk factors (Table 3).

Table 3 Risk factors for preterm delivery in 203 women with pre-existing diabetes by univariate and multivariate logistic regression analysis
Full size table

When restricting the population to women not developing preeclampsia, gestational weight gain was still positively associated with preterm delivery (p = 0.03).

Kidney involvement in early pregnancy (OR 12.71 [3.00, 53.79]), gestational weight gain (100 g/week, OR 1.25 [1.02, 1.54]), estimated fetal weight deviation at 27 gestational weeks (% from the mean, OR 1.07 [1.03, 1.12]) and preeclampsia (OR 7.04 [2.34, 21.19]) were all independently and positively associated with preterm delivery (Table 3). When restricting the variables in the multiple regression analysis to baseline values, the independent predicting factors for preterm delivery were type 1 diabetes and pre-existing kidney involvement (data not shown).

When stratified by diabetes type, 33% of women with type 1 diabetes delivered preterm, compared with 20% of women with type 2 diabetes (Table 1). Preeclampsia leading to preterm delivery occurred with similar prevalence regardless of diabetes type.

In women prescribed prophylactic aspirin, 37% (19 out of 51 women) delivered preterm and 18% developed preeclampsia (9 out of 51 women), compared with 24% (36 out of 152 women) delivering preterm (p = 0.06) and 9% (13 out of 152 women) developing preeclampsia (p = 0.07) in women who were not prescribed aspirin, respectively.

In total, 69% of preterm deliveries were indicated, and the remaining 31% were with spontaneous onset (Table 4). Several contributing indications for preterm delivery were often present concomitantly. The most important causes leading to indicated preterm delivery were suspected fetal asphyxia (n = 17), preeclampsia (n = 12) and fetal overgrowth (n = 5) (Table 4). Falling insulin requirement of 35% (21% to 50%) contributed to suspected fetal asphyxia in ten (59%) of these 17 cases (Table 4). None of these ten women were diagnosed with preeclampsia. In women where preeclampsia was regarded as the main cause of indicated preterm delivery, falling insulin requirements were observed in some of the women, but this was not systematically reported.

Table 4 The clinically most important causes and indications leading to preterm delivery (<37 weeks) in 55 women with pre-existing diabetes. More than one indication was present in most of the cases
Full size table

Discussion

In this prospective observational cohort study of 203 pregnant women with pre-existing diabetes, preterm delivery occurred in around one out of four women with pre-existing diabetes. Preeclampsia, fetal overgrowth and gestational weight gain were independent and potentially modifiable risk factors for preterm delivery. The main causes of indicated preterm delivery were suspected fetal asphyxia and preeclampsia.

The rate of preterm delivery in this study was comparable to previous studies in women with pre-existing diabetes [7, 8, 14] and approximately four times higher than in the Danish background population [6]. Preeclampsia and pre-existing kidney involvement were strong predictors of preterm delivering, in agreement with previous studies [12, 20, 32]. The prevalence of preeclampsia in 26% of women delivering preterm was comparable to the 28% in a French cohort of 441 women with type 1 diabetes [14].

Preterm delivery because of preeclampsia may be reduced by prophylactic aspirin and antihypertensive treatment [13, 17]. It is well described that poor glycaemic control is related to both the onset of preeclampsia and preterm delivery, especially if HbA1c >53 mmol/mol (7%) [7, 9, 11, 14, 33]. In our cohort, HbA1c was close to target in both early and late pregnancy with the majority obtaining HbA1c <53 mmol/mol before delivery, which may explain why elevated HbA1c was not associated with preterm delivery. Larger studies in women with a less optimal glycaemic control may give a more detailed explanation of the association between glycaemic control and the risk of preterm delivery. Improvement in HbA1c levels during pregnancy was most prominent among women with type 2 diabetes and probably due to a lower risk of hypoglycaemia compared with women with type 1 diabetes. According to local guidelines, only women with risk factors for preeclampsia, and therefore also an additionally increased risk of preterm delivery, were recommended prophylactic aspirin in the present study. Therefore, it can be expected that both preterm delivery and development of preeclampsia were seen more commonly in women prescribed prophylactic aspirin. Whether prophylactic aspirin recommended to all pregnant women with pre-existing diabetes, and not restricted to women with additional risk factors, may reduce the prevalence of preeclampsia and preterm delivery needs to be investigated in future studies. Previous cohort studies from our centre suggested that tight antihypertensive treatment results in a reduction in the prevalence of early preterm delivery in women with pre-existing kidney involvement [20, 34]. Tight antihypertensive treatment was given using the same principles and to the same extent for all women in the present study. Nevertheless, further studies evaluating whether strict antihypertensive treatment given regardless of presence of kidney involvement can reduce the development of preeclampsia leading to preterm delivery are warranted.

Gestational weight gain was identified as an independent predictor of preterm delivery in our cohort. This was most prominent in women with type 2 diabetes and in accordance with a register study on 2310 overweight women with type 2 diabetes, where excessive gestational weight gain was associated with preterm delivery [16]. Whether this is also the case in type 1 diabetes needs to be investigated in further studies.

After excluding women with preeclampsia from the dataset, gestational weight gain was still positively associated with preterm delivery, indicating that the association is also present in women not developing fluid retention as part of preeclampsia [22]. Whether interventions to restrict gestational weight gain, such as motivational interviewing to improve adherence to lifestyle recommendations [35], or treatment with metformin without negative impact on the later life of the offspring [36], might reduce preterm delivery needs to be explored in future studies.

It is novel that fetal overgrowth observed as early as 27 gestational weeks was a predictor of preterm delivery. At 33 gestational weeks this signal attenuated, partly because some women already had delivered preterm and partly because the prevalence of fetal overgrowth became high in both groups. Future studies should investigate potential predictors of fetal overgrowth at 27 gestational weeks, i.e. gestational weight gain, HbA1c and additional glucose data including intermittent high glucose spikes [37]. Birthweight SD score in women with type 2 diabetes were numerically more appropriate than in women with type 1 diabetes, as previously seen in the literature [38].

The majority of preterm deliveries in the present study were indicated, which is similar to previous studies in women with type 1 diabetes [7, 8]. Our findings of preeclampsia as a main cause of indicated preterm delivery is also in accordance with previous studies [7, 8]. Nulliparity, fetal growth restriction, oligohydramnios and placental abruption as possible causes of preterm delivery in addition to suspected asphyxia have been reported [7, 8], but this was not seen in our cohort.

Falling insulin requirement was an important component of suspected fetal asphyxia leading to preterm delivery, in line with a previous study where women with type 1 diabetes and a falling insulin requirement >15% were more likely to deliver preterm [15].

Falling insulin requirements have been associated with both preeclampsia and preterm delivery [15, 39]. In the present study, none of the ten women with falling insulin requirements as the main indication leading to indicated preterm delivery were diagnosed with preeclampsia. However, in women where preeclampsia was regarded as the main cause of indicated preterm delivery, falling insulin requirements were also observed in some of the women, but unfortunately not systematically reported. The role of falling insulin requirements leading to preterm deliveries calls for further investigation.

Strengths of our study are the prospectively collected, large cohort of mainly unselected women with pre-existing diabetes, where a combination of clinically relevant endocrine and obstetric risk factors were included, enabling independent modifiable predictors to be identified. In addition, it is novel that the ultrasound fetal size assessments and the causes of preterm delivery were specified.

Despite the cohort originating from a well-defined geographical area roughly covering half of the population in Denmark, it is a limitation that the women were from a single centre with expertise in managing pregnant women with diabetes. This may limit the generalisability of the results to other populations where glycaemic control, weight management and BP control might be different. For ethical reasons, the women had to understand and read Danish to be included. Despite this limitation, the cohort included 24% women of origin other than North-European, and ethnicity was not related to preterm delivery. We used self-reported pre-pregnancy weight to calculate the total gestational weight gain, and therefore there was a risk of recall bias. However, we have previously demonstrated a strong correlation between pre-pregnancy weight and measured weight at first antenatal visit in women with type 2 diabetes [25]. Even though a large number of women within this patient group were included in the study, the number of women with preterm delivery was small, causing a potential lack of statistical power and limiting the possibility of performing multiple regression analysis in women with type 1 and type 2 diabetes separately.

In conclusion, presence of preeclampsia, higher positive ultrasound estimated fetal weight deviation at 27 gestational weeks and higher gestational weight gain were independent potentially modifiable risk factors for preterm delivery in this cohort of women with pre-existing diabetes. Indicated preterm delivery was common with suspected fetal asphyxia or preeclampsia as the most prevalent causes. Prospective studies evaluating whether modifying these predictors will reduce the prevalence of preterm delivery are warranted.