Neonatal outcomes of very low birthweight infants born to mothers with hyperglycaemia in pregnancy: a retrospective cohort study in Japan.

OBJECTIVE: To examine the mortality and morbidities of very low birthweight (VLBW, <1500 g) infants of mothers with hyperglycaemia in pregnancy. DESIGN AND
SETTING: We conducted a retrospective cohort study using data from the Neonatal Research Network of Japan, a nationwide registry of VLBW infants (2003-2012). PATIENTS: We studied 29 626 infants born at 23 to 32 weeks without major congenital anomalies, of which 682 (2.3%) infants were from pregnancies affected by maternal hyperglycaemia. MAIN OUTCOME MEASURES: The primary outcome was hospital mortality. Secondary outcomes were neonatal morbidities and their anthropometric values. Associations between maternal hyperglycaemia and each outcome were observed for the overall period, and statistical tests for interaction were conducted to assess whether they differed before or after the adoption of the International Association of Diabetes in Pregnancy Study Group (IADPSG) guidelines in 2010 for the diagnosis of gestational diabetes mellitus.
RESULTS: Overall, hospital mortality (4.1% vs 5.2%), composite outcomes of mortality and severe morbidity (54.2% vs 60%), and anthropometric values were not significantly different between infants of mothers with or without hyperglycaemia in pregnancy. However, the incidence of respiratory distress syndrome (RDS) in VLBW infants from mothers with hyperglycaemia was significantly higher than those from mothers without it only before (relative risk (RR) 1.09, 95% CI 1.00 to 1.19) and not after (RR 0.97, 95% CI 0.83 to 1.11) the adoption of the IADPSG guidelines.
CONCLUSION: VLBW infants born to mothers with hyperglycaemia in pregnancy do not seem to be at higher risk of mortality and morbidities, except for RDS only before the adoption of the IADPSG guidelines.

The incidence of respiratory distress syndrome (RDS) in term infants from mothers with hyperglycaemia in pregnancy is greater than infants from those not affected by it.

A few studies have reported the mortality and morbidities of very low birthweight (VLBW) infants from mothers with hyperglycaemia in pregnancy.

No study has observed how change in the diagnostic criteria of gestational diabetes mellitus influences the association between maternal hyperglycaemia and mortality and morbidities of VLBW infants.

As a whole, no significant difference was observed in mortality, severe morbidity or anthropometric values between VLBW infants with and without maternal hyperglycaemia in pregnancy.

VLBW infants born to mothers with hyperglycaemia in pregnancy showed a greater risk of RDS only before the adoption of the International Association of Diabetes in Pregnancy Study Group guidelines.

Introduction

Maternal hyperglycaemia in pregnancy is a major complication, known to lead to other serious complications such as intrauterine fetal death, fetal macrosomia, neonatal hypoglycaemia, jaundice, polycythaemia, hypocalcaemia1–3 and preterm birth.4 Depending on the population studied and the diagnostic tests employed, this complication is reported to occur in 2% to 25% of all pregnancies.5

The risk of preterm birth in mothers with diabetes is several times higher than that in mothers without diabetes,6 and very preterm births themselves have a high risk of death and severe morbidity in the first place.7

Furthermore, the effect of diabetes on the child’s health may differ by the severity of the diabetes and the impact of glycaemic control mode on infant outcomes.8 There have been only a few studies to examine the association between hyperglycaemia in pregnancy and the mortality and morbidities of the premature infants.8–12 Several studies had reported that very low birthweight (VLBW, <1500 g) infants born to mothers with hyperglycaemia in pregnancy were at higher risk of necrotising enterocolitis (NEC),8 11 though others had shown not.9 10 12 A recent study conducted in seven countries (including Japan) suggested that hyperglycaemia in pregnancy is not associated with an increased risk of in-hospital mortality or severe morbidity in VLBW infants.12 However, the population combined those diagnosed with gestational diabetes mellitus (GDM) from both before (2007–2010) and after (2010–2015) the publication of the International Association of Diabetes in Pregnancy Study Group (IADPSG) criteria for GDM in 2010. The influence of the change in criteria for diagnosis is non-negligible; for instance, in Japan, after the adoption of the IADPSG criteria for GDM in 2010,13 which resulted in relaxation of the criteria for diagnosis, frequency of GDM pregnancies increased twofold to fourfold to 6%–12%.14 15

Thus, the purpose of this study was to exsamine the association between maternal hyperglycaemia in pregnancy and mortality, morbidities, birth weight for gestational age and extrauterine growth of VLBW infants and investigate whether this association differed before and after the adoption of the IADPSG criteria for GDM diagnosis. Japan has one of the lowest neonatal mortality and morbidity rates among the developed countries,16 and the findings of this study would be useful internationally to understand the impact of maternal hyperglycaemia in pregnancy on preterm infants in an advanced perinatal care.

Methods

Study population

Patient data from the Neonatal Research Network of Japan (NRNJ), which contain prospectively registered data on VLBW infants admitted to the 204 participating neonatal intensive care units (NICUs) nationwide were analysed.17 All institutes registered individual data, including clinical characteristics and morbidities, for the VLBW infants under their care. Data from the period of 1 January 2003 to 31 December 2012 were examined. The data consisted of VLBW infants from 96% (72/75 as of 2008) of all level 3 NICUs and from seven level 2 NICUs in Japan.18

Infants born before 23 weeks and after 33 weeks were excluded from the analysis, as the former have a high mortality rate and almost all of the latter were small for gestational age (SGA), which has a strong impact on mortality and morbidity. We also excluded VLBW infants born between 23 and 32 weeks with congenital anomalies (serious congenital heart disease or major genetic disorder), those who died at the delivery room, and those with missing data on maternal diabetes, or missing data on their characteristics and outcomes.

Definition of disease and outcomes

For each infant, the NRNJ database collects data on maternal background (including age, parity) and infant outcomes (including birth anthropometrics, whether or not specific complications had developed during hospitalisation). A database manual was distributed to each hospital to ensure the definition of variables was uniform across the participating institutions.

Neonatal morbidities recorded in the database included: respiratory distress syndrome (RDS), patent ductus arteriosus (PDA), necrotising enterocolitis (NEC), intraventricular haemorrhage (IVH), periventricular leukomalacia (PVL), chronic lung disease (CLD) (measured at 28 days of age and at 36 weeks) and retinopathy of prematurity (ROP). From these data, we defined the composite outcome as the infant experiencing any of the following: hospital death, requiring PDA treatment (indomethacin or ligation), NEC, IVH grades III–IV, PVL, CLD at 36 weeks or requiring ROP treatment (laser photocoagulation or antivascular endothelial growth factor).

The diagnosis criteria of all complications were predefined in the distributed manual as follows: RDS was diagnosed based on clinical presentation and radiographic findings; PDA was diagnosed based on both echocardiographic and clinical findings; NEC was defined as Bell stage II or greater; IVH was diagnosed based on cranial ultrasound and graded using the classification of Papile et al 19; PVL was diagnosed based on cranial ultrasound or MRI under the presence of cystic lesion around periventricular white matter; CLD 28 and CLD 36 were defined as the need for supplemental oxygen at 28 days of age or at 36 weeks of postmenstrual age, respectively; and ROP was staged according to the classification defined by the Japanese Ministry of Health, Labour and Welfare, which directly correlates with ICROP (International Classification of Retinopathy of Prematurity).

From the birth and discharge anthropometrics included in the database, z-scores for infant weight, height and head circumference were calculated using neonatal anthropometric charts—reference data of vaginal deliveries during 2003–2005 in Japan.20 21 SGA was defined as birth weight being less than the 10th percentile for the gestational age based on the same reference data. Change in z-scores for each measurement from birth to discharge was calculated by subtracting the z-scores from each other.

Data on the presence of glucose intolerance during pregnancy were collected throughout the period, as a yes/no question. In Japan, the random blood glucose test is used for screening of GDM during the first trimester. Then, between 24 and 28 weeks of gestation, either the random blood glucose or glucose challenge test (50 g) were performed for screening. These protocols did not change throughout the study period. However, in 2010, GDM diagnosis criteria changed in Japan, when the Japan Society of Obstetrics and Gynecology (JSOG) adopted the IADPSG criteria.13 Before this adoption, diagnosis of GDM was determined according to the JSOG criteria, in which two or more abnormal values of the following were required in the 75 g oral glucose tolerance test (OGTT): fasting blood glucose ≥100 mg/dL (≥5.6 mmol/L); 1 hour ≥180 mg/dL (≥10.0 mmol/L) or 2 hours ≥150 mg/dL (≥8.3 mmol/L). After adopting the IADPSG criteria, diagnosis of GDM was based on one of the following glucose levels in the 75 g OGTT: fasting blood glucose ≥92 mg/dL (≥5.1 mmol/L); 1 hour ≥180 mg/dL (≥10.0 mmol/L) or 2 hours ≥153 mg/dL (≥8.5 mmol/L). A previous study showed that this change increased the prevalence of GDM.14 Thus, 2003–2009 was defined as the pre-IADPSG phase and 2011–2012 as the post-IADPSG phase, and we categorised the infants based on birth year accordingly. Data on the presence of maternal glucose intolerance during pregnancy were recorded except for all of the following items: subtype of diabetes mellitus (DM) (GDM, overt DM in pregnancy or pre-existing DM), status of glycaemic control and presence of treatment and its content.

Patient involvement

Patients were not directly involved in the design of this study.

Statistical analysis

First, the background characteristics of the mothers and infants were described using proportions for categorical variables and means with SD for continuous variables. Categorical variables were compared with χ2 test and continuous variables with two-sample t-test.

Second, to evaluate the association between hyperglycaemia in pregnancy and dichotomous neonatal outcomes, Poisson regression analyses were performed to calculate adjusted relative risk (RR) and 95% CIs. Multivariate linear regression analyses were performed to calculate coefficients and 95% CIs for continuous outcomes. For these analyses, based on clinical relevance, the following variables were introduced into the models as confounders: maternal age, parity, mode of delivery, use of antenatal steroid, sex of the infants, gestational week, SGA and Apgar score at 5 min. Maternal ages were left continuous and investigated non-linearities using cubic spline regression model. As a large number of anthropometric values at discharge were missing (n=9754, 11 446, 11 454 for weight, height and head circumference, respectively), multiple imputation methods were carried out with chained equations to impute the missing values with exclusion of missing data from infants discharged after 41 weeks of gestational age which were unsuitable for predicting anthropometric z-scores accurately. Complete case analysis, as well as analyses not including variables which could act as mediators (SGA, Apgar score at 5 min) was also done as sensitivity analysis.

Third, to assess whether the association differed between the pre-IADPSG and post-IADPSG phases, statistical tests for interaction between the main effect and time period were conducted using multivariate logistic regression analysis of each outcome, and stratified analyses were conducted if significant or close to significant.

In all analyses, a two-sided p-value <0.05 was considered statistically significant with the exception of p-value for interaction <0.1. Stata V.14.0 (College Station, Texas, USA) was used for all analyses.

Ethics statement

This study was approved by the Internal Review Board of Tokyo Women’s Medical University, where all data were collected and stored. Written informed consent was obtained from the parents or guardians.

Results

Figure 1 shows the population of this study. The NRNJ cohort was composed of 40 806 VLBW infants born during the study period. Of these, 34 977 infants were born at 23 to 32 weeks of gestation. Of these, 506 infants were excluded because of missing data on maternal diabetes, 99 infants were excluded because of death at delivery room, 1816 infants were excluded because of congenital anomaly and 2930 infants were excluded due to missing data on their characteristics and outcomes. Finally, 29 626 infants were included in the analysis.

Figure 1

Population flow chart. DM, diabetes mellitus; GA, gestational age.

The characteristics of the mothers and infants included in this study are shown in table 1. Mothers with hyperglycaemia in pregnancy was significantly older than mothers without it (p<0.001). Gestational week and median birth weight in infants of mothers with hyperglycaemia in pregnancy were significantly higher than that of mothers without it (p=0.03, p<0.01, respectively). Values of weight and length at discharge in infants of mothers with hyperglycaemia in pregnancy were higher than those of mothers without it (p<0.01, p<0.05, respectively) except for head circumference (p=0.53). The proportions of singleton, caesarean delivery, use of antenatal steroids, male (sex), and SGA, and Apgar scores were not significantly different between the two groups.

Table 1

Maternal and infant characteristics of 29 626 very low birthweight infants

Characteristics	Infants with maternal hyperglycaemia in pregnancy (N=682)	Infants without maternal hyperglycaemia in pregnancy (N=28 944)	P value
Maternal
Maternal age (SD)	33.8±4.8	31.2±5.8	<0.01
Maternal age, missing, n (%)	8 (1.2)	803 (2.8)
Singleton, n (%)	501 (73.5)	21 652 (74.5)	0.42
Caesarean delivery, n (%)	552 (80.9)	22 694 (78.4)	0.11
Antenatal steroids, n (%)	318 (46.6)	14 223 (49.1)	0.20
Neonatal
Male, n (%)	375 (55.0)	14 896 (51.4)	0.07
SGA, n (%)	222 (32.6)	9458 (32.7)	0.33
Apgar score at 1 min, mean (SD)	5.3±2.4	5.3±2.4	0.49
Apgar score at 5 min, mean (SD)	7.4±1.9	7.3±1.9	0.20
Apgar score at 5 min <7, n (%)	164 (24.1)	7588 (26.2)	0.20
Gestational age (weeks), mean (SD)	28.1±2.5	27.9±2.5	0.03
Birth weight (g)	1045±286	1010±293	<0.01
Measurement at discharge
Weight (g)	2798±880 (n=670)	2716±850 (n=27 406)	<0.01
Length (cm)	46.6±4.9 (n=633)	46.2±4.9 (n=25 503)	<0.05
Head circumference (cm)	34.3±3.1 (n=641)	34.2±3.1 (n=25 616)	0.53

SGA, small for gestational age.

Table 2 compares the mortality and morbidities between the two groups. There was no significant difference between the two groups in regard to any morbidity either in the crude analysis or after adjusting for confounders as previously described.

Table 2

Mortality and morbidities among infants with or without maternal hyperglycaemia in pregnancy

Outcomes, n (%)	Infants with maternal hyperglycaemia in pregnancy (n=682)	Infants without maternal hyperglycaemia in pregnancy (n=28 944)	P value	Adjusted* RR (95% CI)
Hospital death	28 (4.1)	1504 (5.2)	0.20	0.83 (0.58 to 1.20)
RDS	438 (64.2)	17 842 (61.6)	0.17	1.04 (0.99 to 1.1)
NEC	9 (1.5)	440 (1.5)	0.67	0.92 (0.48 to 1.77)
PDA required any treatment	250 (36.7)	10 964 (37.9)	0.52	1.00 (0.92 to 1.1)
IVH grades III–IV or PVL	47 (6.9)	2328 (8.0)	0.27	0.86 (0.65 to 1.15)
CLD 28	236 (34.6)	10 822 (37.4)	0.13	0.97 (0.89 to 1.06)
ROP required any treatment	101 (14.8)	4662 (16.1)	0.36	0.99 (0.84 to 1.18)
Composite†	370 (54.2)	16 205 (60.0)	0.37	1.00 (0.95 to 1.07)

*Adjusted for maternal age, parity, mode of delivery, sex, antenatal steroid use, gestational week, SGA and Apgar score at 5 min.

†Includes hospital death, NEC, PDA required any treatment, IVH or PVL, CLD 36 and ROP required any treatment.

CLD, chronic lung disease; IVH, intraventricular hemorrhage; NEC, necrotizing enterocolitis; PDA, patent ductus arteriosus; PVL, periventricular leukomalacia; RDS, respiratory distress syndrome; ROP, retinopathy of prematurity.

Statistical association between hyperglycaemia in pregnancy and incidence of RDS and ROP of the infants was almost close to significant between infants born in the pre-IADPSG and post-IADPSG phase (p-value for interaction=0.10, 0.12, respectively) (table 3). For both outcomes, stratified analyses by study period were performed. As a result, for those born in the pre-IADPSG phase, infants of mothers with hyperglycaemia in pregnancy had higher rates of RDS (RR 1.09, 95% CI 1.00 to 1.19) than infants of mothers without it. This difference was largely affected by infants born at 28 to 29 weeks of gestation (online supplementary figure). For those born in the post-IADPSG phase, there was no significant difference in the incidence of RDS between infants of mothers with and without hyperglycaemia in pregnancy (RR 0.97, 95% CI 0.83 to 1.11). Regarding the incidence of ROP, there was no significant difference for those born in the pre-IADPSG and post-IADPSG phase (RR 1.11, 95% CI 0.87 to 1.42 and RR 0.85, 95% CI 0.64 to 1.11, respectively).

Table 3

Interaction between maternal hyperglycaemia in pregnancy and periods (pre-IADPSG and post-IADPSG) on several neonatal outcomes

Outcomes	P-value for interaction*†
Hospital death	0.61
RDS	0.1
NEC	0.88
PDA required any treatment	0.68
IVH or PVL	0.19
CLD 28	0.61
ROP required any treatment	0.12
Composite‡	0.80

*Calculated using multivariate logistic regression analysis with adjustment for maternal age, parity, mode of delivery, sex, antenatal steroid use, gestational week, SGA, Apgar score at 5 min.

†P value for interaction <0.1 is considered significant.

‡Includes hospital death, NEC, PDA required any treatment, IVH or PVL, CLD 36 and ROP requiring any treatment.

CLD, chronic lung disease; IVH, intraventricular hemorrhage; NEC, necrotizing enterocolitis; PDA, patent ductus arteriosus; PVL, periventricular leukomalacia; RDS, respiratory distress syndrome; ROP, retinopathy of prematurity.

Table 4 shows the distribution of anthropometric z-scores for infants from mothers with and without hyperglycaemia in pregnancy, after multiple imputations. Z-scores of weight, height and head circumference, as well as change in z-scores for all three measures were not significantly different between the two groups. The sensitivity analyses, when using complete cases (and not conducting multiple imputations) of the results were not significantly different in all measurements.

Table 4

Growth measure z-scores among infants with or without maternal hyperglycaemia in pregnancy (multiple imputation used to impute missing values n=29 626)

	Infants with maternal hyperglycaemia in pregnancy	Infants without maternal hyperglycaemia in pregnancy	P value*
Weight z-score at discharge	−1.26	−1.39	0.19
Change in weight z-score	−0.58	−0.65	0.2
Length z-score at discharge	−1.54	−1.61	0.35
Change in length z-score	−1.17	−1.17	0.35
Head circumference z-score at discharge	0.39	0.22	0.2
Change in head circumference z-score	0.34	0.26	0.49

*Multivariate logistic regression analysis with adjustment for maternal age, parity, mode of delivery, sex, antenatal steroid use, gestational week and SGA.

SGA, small for gestational age.

Discussion

In this study, we examined the association between maternal hyperglycaemia in pregnancy and mortality and morbidities of VLBW infants based on a large, nationwide Japanese database. The present study demonstrated that maternal hyperglycaemia in pregnancy has no impact on the outcomes for VLBW infants except for RDS. Furthermore, the increase in risk of RDS was only significant for pregnancies with maternal hyperglycaemia in pregnancy diagnosed according to the previous guidelines.

To the best of our knowledge, to date, five studies have examined the association between maternal hyperglycaemia in pregnancy and mortality and morbidities of VLBW infants. Of the five studies, three reported that maternal hyperglycaemia in pregnancy was unrelated to any major complications for VLBW infants.9 10 12 On the other hand, the other two studies showed that VLBW infants born to mothers with hyperglycaemia in pregnancy had a higher incidence of NEC.8 11 In the most recent and largest study to date, Persson et al 12 reported that the risk of a variety of adverse outcomes (NEC, bronchopulmonary dysplasia, ROP treatment, treated PDA, IVH grades III–IV, cystic PVL, RDS and in-hospital mortality) or their composite outcome did not differ between infants of mothers with or without diabetes. With the exception of RDS, these insignificant findings might be acceptable since associations between other severe morbidities and maternal hyperglycaemia were not universally noted.

It is generally believed that inadequate control of hyperglycaemia in pregnancy exposes the infant to the risk of RDS, as insulin inhibits gene expressions of surfactant proteins A and B in lung epithelial cells.22–24 Mature infants born to mothers with diabetes has shown to carry a sixfold increase in risk of RDS, consistent with this hypothesis.25 However, all five previous studies on VLBW failed to find a significant association between maternal hyperglycaemia in pregnancy and the risk of RDS. Our study is the first to demonstrate that the incidence of RDS increased with maternal hyperglycaemia in pregnancy, but this association was present only in those born before the diagnostic criteria of GDM was changed in 2010.

Our finding that increased risk of RDS was observed only in infants of mothers with hyperglycaemia in pregnancy diagnosed before the relaxation of the GDM diagnostic criteria in 2010 suggests that only the more severe cases of maternal hyperglycaemia in pregnancy carry higher risk of RDS in infants. This finding is consistent with previous reports on mature infants, where RDS risk was found to be highest among those with the most severe cases of diabetes, namely, unstable type 1 diabetes.25 Our study may have had the power to detect the association between hyperglycaemia in pregnancy and RDS due to its relatively large sample size and our homogeneous definition of GDM used. In comparison, a recent large-scale analysis, conducted across seven countries, which failed to detect this association, had inconsistencies in its diagnostic criteria to assess GDM.

Interestingly, although about half of the infants in this study did not receive antenatal steroids which may help mature the lung, the rate of RDS among infants of mothers with diabetes (64%) was comparable to other populations where only about 20% did not receive antenatal steroids.9–11 While this is only a speculation, this discrepancy in the incidence of RDS may be attributable to the difference in risk of RDS by race: in more mature infants (34 to 42 weeks of gestational age), Asian infants have been reported to have a lower risk for RDS than Caucasian infants.26

It is universally known that infants born at term from mothers with diabetes tend to have larger birth weights (unless the disease is severe); however, anthropometric data of preterm infants of maternal diabetes is lacking. In our study, birth weight in VLBW infants of maternal hyperglycaemia in pregnancy was significantly higher than those in the control group. Our study findings differed from those from Boghossian who reported that extremely preterm infants of mothers using insulin before pregnancy were smaller at birth (height and head circumference, but not weight) than mothers who do not use insulin before pregnancy, which the author attributed to prolonged hyperglycaemia causing deterioration of vascular condition.8 Most of the macrosomia of infants from mothers with diabetes develops in the third trimester, when maternal lipids as well as glucose produce increased fetal adiposity. Comparison of Boghossian’s study and ours is difficult due to the paucity of detailed information on the types of DM and treatment in our study. One possible explanation to the difference in findings may be differences in prevalence of insulin therapy. Although we do not have the exact proportion, considering the general prevalence of pregestational DM in Japan (<3%), it is likely that most women included in our study were not receiving insulin treatment prior to pregnancy. Our findings generate a hypothesis that mothers with hyperglycaemia in pregnancy who do not require insulin before pregnancy deliver larger weight infants even if the delivery is very preterm, while those with more severe diabetes, who require the use of insulin before pregnancy, the infants tend to be smaller. Further research should be conducted on this topic.

We acknowledge that there are several limitations to this study. First, we did not have data on the type and onset of DM, status of glycaemic control, presence and details of treatment. Furthermore, the mothers were assessed their glycaemic status at one time point between 24 and 28 weeks of gestation, and thus we were not able to account for the increase in risks of hyperglycaemia in proportion to its duration. These paucity of information on maternal diabetes might lead to insufficient assessment of some risks such as NEC,8 in which Boghossian reported that preterm infants born to mothers with insulin use before pregnancy had higher risk of NEC than infants born to mothers with insulin use started during pregnancy and without insulin use. Second, our study population was limited to preterm infants admitted to the NICU, thus excluding delivery room deaths, infants with congenital anomalies and infants born at term. Thus, we were not able to evaluate disorders of infants of mothers with diabetes, such as hyperbilirubinaemia, polycythaemia, hypocalcaemia and asymmetric cardiac septal hypertrophy. Third, our study had a high proportion of missing data on anthropometric values at discharge. Although we conducted imputation to account for the missing data, we cannot rule out the chance of bias.

Despite these limitations, our findings can be interpreted as revealing the overall trend in regard to results of VLBW infants born to mothers with hyperglycaemia in pregnancy, as in Asian women, GDM accounts for the majority of occurrences of hyperglycaemia in pregnancy.27

In conclusion, from our study, which included a large number of VLBW infants of mothers with hyperglycaemia in pregnancy, we found that there is no relation between hyperglycaemia in pregnancy and in-hospital mortality and risk of severe morbidities. Nevertheless an association between risk of RDS and hyperglycaemia in pregnancy differed according to the criteria used for GDM diagnosis. We found that before the IADPSG guidelines for GDM diagnosis was adopted in 2010, VLBW infants born to mothers with hyperglycaemia in pregnancy had an increased risk of RDS.