Maternal lipids associated with large-for-gestational-age birth weight in women with type 1 diabetes: results from a prospective single-center study.

INTRODUCTION: Despite improvement in diabetes care over the years, the incidence of macrosomia in type 1 diabetic mothers is still very high and even shows an increasing tendency. It is suggested that other factors that maternal hyperglycemia might be associated with excessive fetal growth in diabetic mothers. The aim of this study was to determine whether maternal lipids might contribute to high rates of large-for-gestational-age (LGA) newborns in women with type 1 diabetes (T1DM).
MATERIAL AND METHODS: This prospective, single-center study was performed in a population of women with T1DM admitted to the perinatal center for women with diabetes. Data were collected in the first trimester (< 12th week), in mid-pregnancy (20th-24th weeks), and before delivery (34th-39th weeks).
RESULTS: Among 114 women included in the analysis, 30 (26.3%) delivered LGA newborns. The remaining 84 (73.7%) newborns were appropriate for gestational age (AGA). Lower high-density lipoprotein (HDL) HDL concentration in the first trimester was significantly associated with LGA (p = 0.01). Similar associations were observed for the HDL concentrations in mid-pregnancy (p = 0.04) and before delivery (p = 0.03). Higher triglyceride concentrations in the first trimester (p = 0.02) and before delivery (p = 0.008) were associated with LGA. Higher glycated haemoglobin (HbA1c) levels in mid-pregnancy and before delivery were associated with LGA. The associations between maternal lipids and LGA were independent of maternal body mass index at onset of the study, gestational weight gain and HbA1c concentrations.
CONCLUSIONS: Decreased HDL and increased triglycerides during pregnancy might contribute to the development of LGA in women with type 1 diabetes.

Introduction

Excessive fetal weight, which is defined as large for gestational age (LGA), is one of the most significant complications of pregnancy in women with type 1 diabetes. Despite improvement in diabetes care over the years, the incidence of LGA in type 1 diabetic mothers is still very high and even shows an increasing tendency [1-3]. This is worrying as LGA is associated with a high risk of maternal and fetal complications [4, 5]. The importance of prevention of LGA is not only an issue in perinatal medicine as LGA infants are at increased future risk of developing obesity, diabetes, and cardiovascular disease due to fetal programming [6-11].

The association between maternal hyperglycemia and LGA has been confirmed in a number of studies [12-14]. Therefore, preventive strategies against excessive fetal growth have been focused on achieving and maintaining near normal glycemic levels throughout pregnancy. However, such an approach was proved to be not fully effective in many women with type 1 diabetes whose fetuses develop LGA despite good glycemic control [15].

Some other factors have been previously linked with LGA, including maternal overweight/obesity, excessive gestational weight gain, and lipids. The first two factors have also been studied in women with type 1 diabetes, but the association between lipids and LGA in this population has not been extensively investigated [16-18]. There is one study in the literature showing that increased levels of triglycerides and decreased levels of high-density lipoprotein (HDL) in the third trimester might be linked to LGA in patients with pregestational diabetes. However, the results of this study cannot be directly extrapolated to women with type 1 diabetes as the analyzed group consisted of both type 1 and type 2 diabetics [19]. Much more data on lipids and LGA come from normal pregnancies and pregnancies complicated by gestational diabetes mellitus (GDM). In non-diabetic patients LGA was related to increased triglycerides, increased free fatty acids, and decreased HDL during early pregnancy [20]. In women with GDM similar findings were reported in mid-pregnancy and near delivery, except that a relationship with HDL was not demonstrated [21-23].

Moreover, recent data suggest that maternal lipid profile in early gestation is an independent factor related to childhood adiposity at the age of 5 to 6 years in non-diabetics [24].

The aim of our study was to determine whether changes in maternal lipids analyzed at three different time points during pregnancy (in the first trimester, in mid-pregnancy, and before delivery) are associated with LGA in women with type 1 diabetes.

Material and methods

Study setting

This study was conducted in the Department of Obstetrics and Women’s Diseases, Gynecologic and Obstetrical University Hospital in Poznan, Poland. The department with its outpatient clinic is the biggest perinatal center for pregnant women with diabetes in Poland, providing care for patients from the Greater Poland province (population of approximately 3.4 million). According to Polish Diabetes Association recommendations and our internal standards, every patient with preexisting diabetes from Greater Poland is immediately referred to our department once pregnancy is confirmed. The process of care delivered to diabetic women without complications is based on at least 3 planned, short-stay hospital admissions during pregnancy: in the first trimester, in mid-pregnancy (20th–24th weeks of gestation), and near delivery (34th–35th weeks of gestation). Patients who require more vigilant surveillance are admitted more frequently. In between hospital admissions patients are referred biweekly for regular check-ups in the hospital-based outpatient clinic.

Study design and patients

This prospective study was conducted between June 2012 and December 2014. The study group represents a population of pregnant women with type 1 diabetes admitted to the Department of Obstetrics and Women’s Diseases in the first trimester and followed up to delivery.

Patients were all Caucasian and received standard pregnancy care for diabetes as recommended by the Polish Diabetes Association and Polish Gynecological Society, targeting a fasting glucose level of 3.3–5.5 mmol/l, 1-hour postprandial glucose below 6.7 mmol/l, and glycated haemoglobin (HbA1c) below 6.1% (43 mmol/mol) [25].

All women were on intensive insulin therapy using either multiple daily injections (MDI) or continuous subcutaneous insulin infusion (CSII).

Exclusion criteria were multiple pregnancy, miscarriage, preeclampsia, delivery before completed 34th week of gestation, delivery in another hospital, or loss to follow-up.

Anthropometric, clinical, and laboratory data were collected during 3 planned hospitalizations: in the first trimester (< 12th week of gestation), in mid-pregnancy (20th–24th weeks of gestation), and before delivery (34th–39th weeks of gestation).

Blood samples were taken after overnight fasting and immediately transported to the central laboratory of the Gynecologic Obstetrical University Hospital in Poznan for analysis. HbA1c level in whole blood was determined using the turbidimetric inhibition immunoassay (TINIA), Tina-quant Hemoglobin A1c II test in a Cobas c311 analyzer (Roche Diagnostics, Basel, Switzerland). The normal range for this test is 4.8–6.0% (29–42 mmol/mol) for a non-pregnant population.

The total serum cholesterol, HDL cholesterol and triglyceride (TG) levels were determined with Roche Diagnostics reagents (Cholesterol CHOD-PAP, HDL-C plus, and Triglycerides GPO-PAP, respectively) on a Cobas c501 analyzer. The following formula was used to calculate the level of low-density lipoprotein (LDL) cholesterol: LDL cholesterol = total cholesterol – HDL cholesterol – (TG/5).

Anthropometric measurements (height, weight, and waist/hip circumference) and blood pressure measurements were performed at the onset of the study.

Large for gestational age was defined as birth weight greater than the 90th percentile using age- and sex-specific regional growth charts [26].

The Institutional Ethical Committee of Poznan University of Medical Sciences (no. 673/12, June 12, 2012) approved the study protocol. Informed, written consent was obtained from every patient before inclusion in the study.

Statistical analysis

Statistical analyses were performed using MedCalc for Windows, version 12.1.3.0 (MedCalc Software, Mariakerke, Belgium). Testing for normality of data distribution was performed using the D’Agostino-Pearson test. Student’s t-test was used to measure the significance of the difference between two continuous variables when data fitted a normal distribution, with results expressed as mean ± standard deviation (SD). In the case of non-normally distributed data, comparisons were made using the Mann-Whitney test, with results expressed as the median and interquartile range (IQR). The χ2 and Fisher exact tests were used for the comparison of categorical variables. Spearman’s rank correlation coefficient was used to test the relationship between two variables when data did not follow a normal distribution.

Simple logistic regression was used to search for lipid fractions associated with LGA. Those parameters found to be independently associated with LGA were included in multivariate logistic regression (stepwise) models built separately for each follow-up hospitalization.

Based on clinical judgment, we included the following confounders in the stepwise models: maternal age, duration of diabetes, body mass index (BMI) at onset of the study, gestational weight gain, parity (multipara vs. primipara), mode of insulin therapy (CSII vs. MDI), and HbA1c [27].

According to the stepwise method, variables were entered into the model if their associated P-values were less than 0.05 and then sequentially removed if their associated p-values became greater than 0.2. Testing for the goodness of fit of a logistic regression model was performed with the Hosmer-Lemeshow test.

Statistical significance was defined as p < 0.05 (two-sided).

Results

Of the 171 women with type 1 diabetes admitted to our department in the first trimester, 114 (64%) were eligible for the final analysis. Two patients chose not to participate in the study. One patient was excluded from the study group due to twin pregnancy. There were 12 spontaneous abortions and 1 intrauterine fetal death in the third trimester. One woman delivered preterm in the 28th week of gestation. Eleven women were excluded from the analysis due to preeclampsia. Twenty-nine patients were lost to follow-up or delivered in other hospitals.

Among 114 women included in the analysis 30 (26.3%) delivered LGA newborns. The remaining 84 (73.7%) newborns were appropriate for gestational age (AGA). There were no small-for-gestational-age (SGA) newborns in the analyzed group.

Clinical and laboratory characteristics of the AGA and LGA groups are compared in Tables I and II.

Table I

Comparison of clinical characteristics between study groups

Variable	AGA N = 84	LGA N = 30	P-value
Maternal characteristics:
Gestational age at onset of study, mean ± SD [weeks]	8 ±2	9 ±2	0.24
Maternal age at onset of study, median (IQR) [years]	29 ±4	28 ±5	0.23
Primipara, n (%)	49/84 (58.3)	13/30 (43.3)	0.23
Preconception care, n (%)	38/84 (45.2)	13/30 (43.3)	0.97
Duration of diabetes mean ± SD [weeks]	11 ±7	13 ±7	0.11
BMI at onset of study, median (IQR) [kg/m²]	22.6 (20.7–24.8)	23.4 (21.6–26.7)	0.08
Waist-to-hip ratio at onset of study, mean ± SD	0.79 ±0.06	0.79 ±0.06	0.65
Gestational weight gain, mean ± SD [kg]	11.5 ±4.7	12.3 ±3.8	0.4
Mode of insulin therapy during pregnancy, n [MDI/CSII]	20/64	7/23	0.84
Systolic blood pressure at onset of the study, mean ± SD [mm Hg]	115 ±10	114 ±10	0.57
Diastolic blood pressure at onset of the study, mean ± SD [mm Hg]	71 ±7	72 ±8	0.52
Infant characteristics :
Gestational age, median (IQR) [weeks]	38 (37–39)	38 (37–38)	0.67
Birth weight, median (IQR) [g]	3390 (3040–3630)	4035 (3900–4260)	< 0.0001

Table II

Comparison of laboratory characteristics between study groups

Variable	AGA N = 84	LGA N = 30	P-value
HbA1c I, %; median (IQR) [mmol/l]	6.2 (5.8–7.2); 44 (40–55)	6.9 (6.1–7.6); 52 (43–60)	0.1
HbA1c II, %; median (IQR) [mmol/l]	5.4 (5.1–5.7); 36 (32–39)	5.8 (5.6–6.1); 40 (38–43)	0.0002
HbA1c III, %; median (IQR) [mmol/l]	5.7 (5.5–6.2); 39 (37–44)	6.2 (5.9–6.6); 44 (41–49)	0.0015
Serum total cholesterol I, mean ± SD [mmol/l]	4.39 ±0.63	4.23 ±0.62	0.23
Serum total cholesterol II, mean ± SD [mmol/l]	5.88 ±1.01	5.89 ±0.86	0.97
Serum total cholesterol III, mean ± SD [mmol/l]	6.93 ±1.37	6.50 ±1.49	0.22
Serum HDL cholesterol I, median (IQR) [mmol/l]	1.99 (1.70–2.21)	1.64 (1.54–1.87)	0.0012
Serum HDL cholesterol II, median (IQR) [mmol/l]	2.32 (2.06–2.66)	2.17 (1.79–2.50)	0.05
Serum HDL cholesterol III, median (IQR) [mmol/l]	2.08 (1.78–2.45)	1.85 (1.46–2.23)	0.03
Serum LDL cholesterol I, median (IQR) [mmol/l]	2.03 (1.72– 2.45)	2.12 (1.70–2.41)	0.86
Serum LDL cholesterol II, median (IQR) [mmol/l]	2.99 (2.49–3.42)	3.14 (2.56–3.52)	0.38
Serum LDL cholesterol III, median (IQR) [mmol/l]	3.74 (3.00–4.39)	3.18 (2.88–3.95)	0.22
Serum triglycerides I, median (IQR) [mmol/l]	0.62 (0.51–0.77)	0.75 (0.56–0.96)	0.04
Serum triglycerides II, median (IQR) [mmol/l]	1.40 (1.12–1.74)	1.58 (1.11–1.96)	0.27
Serum triglycerides III, median (IQR) [mmol/l]	2.32 (1.97–2.93)	2.61 (2.31–3.55)	0.04

In the first trimester, mothers of LGA newborns had lower HDL (1.64 vs. 1.99, p = 0.0012), higher triglycerides (0.75 vs. 0.62, p = 0.04), and comparable HbA1c (6.9 vs. 6.2, p = 0.1).

In mid-pregnancy, lower HDL persisted in the LGA group (2.17 vs. 2.32, p = 0.05), but there was no significant difference in maternal triglycerides between the groups. Metabolic control of diabetes improved between the first and the second hospital admission in both groups. However, women who delivered LGA newborns had significantly higher HbA1c in mid-pregnancy (5.8 vs. 5.4, p = 0.0002).

Before delivery, lower HDL was still observed in the LGA group (1.85 vs. 2.08, p = 0.03) as well as higher triglycerides (2.61 vs. 2.32, p = 0.04), and HbA1c (6.3 vs. 5.7, p = 0.0015).

We did not find any differences in total cholesterol and LDL during all follow-up hospitalizations. There was no difference in triglycerides in mid-pregnancy between the groups.

Triglycerides were inversely correlated with HDL in the first trimester (rho = –0.303, p = 0.001), in mid-pregnancy (rho = –0.255, p = 0.01), and before delivery (rho = –0.325, p = 0.001).

Prediction of LGA with maternal lipids and HbA1c

In the multivariate logistic regression analysis lower HDL concentration in the first trimester was significantly associated with LGA (Table III).

Table III

Prediction of LGA with maternal lipids (mmol/l) and HbA1c (%) using multivariate logistic regression models*

Variable	Odds ratio (95% CI)	Coefficient	P-value
Serum HDL cholesterol I	0.225 (0.069–0.74)	–1.49	0.01
Serum HDL cholesterol II	0.329 (0.112–0.967)	–1.11	0.04
Serum HDL cholesterol III	0.304 (0.101–0.918)	–0.19	0.03
Serum triglycerides I	5.522 (1.279–23.843)	1.71	0.02
Serum triglycerides III	2.14 (1.128–3.765)	0.76	0.008
HbA1c II	2.279 (1.143–4.541)	0.824	0.01
HbA1c III	1.962 (1.096–3.513)	0.674	0.02

Data shown as adjusted odds ratios (OR) and associated 95% confidence intervals (95% CI). Models adjusted for maternal age, duration of diabetes, BMI at onset of the study, gestational weight gain, parity (multipara = 1 vs. primipara = 0), mode of insulin therapy (CSII = 1 vs. MDI = 0). Models assessing lipids were additionally adjusted for HbA.

A similar association was observed for the HDL concentrations in mid-pregnancy and before delivery.

Higher TG concentration in the first trimester and before delivery was associated with increased risk of LGA. There was no association between second trimester TG and LGA.

Both total cholesterol and LDL in the first trimester, in mid-pregnancy, and before delivery were not associated with LGA.

Higher HbA1c was positively associated with LGA in mid-pregnancy and before delivery, with no association in the first trimester.

Discussion

This study is among the first to evaluate the influence of maternal lipid profile on the risk of fetal excessive growth in women with type 1 diabetes.

We demonstrated that lower HDL in all trimesters and higher triglycerides in the first and third trimesters were associated with LGA in this population.

Such an association has been shown previously, but only in the third trimester, by Göbl et al. in women with pregestational diabetes. The authors did not find any association between either total cholesterol or LDL and LGA, which is in line with our findings [19]. Additionally, their study is not fully representative of solely the population of women with type 1 diabetes, because their study group also included women with type 2 diabetes.

As fetal nutrition depends mainly on glucose, maternal hyperglycemia has been suggested to be the major determinant of abnormal fetal growth in diabetic mothers. However, it has been shown by Evers et al. that many women with type 1 diabetes deliver LGA infants despite relatively good glycemic control. The authors demonstrated that third trimester HbA1c was the most powerful predictor of macrosomia, but its predictive capacity was weak [15].

In our patients the association between HbA1c and the risk of LGA was the strongest in the second trimester, followed by the third trimester. Nonetheless, it should be emphasized that despite statistically higher levels of HbA1c, mothers of LGA infants were relatively well controlled in the second part of pregnancy. This further suggests that not only maternal hyperglycemia might be responsible for abnormal fetal growth in women with type 1 diabetes, which has been shown in our study.

We found that prediction of LGA might be possible early in pregnancy based on the maternal lipid profile, but cannot be based on HbA1c. In our study, many patients who were uncontrolled on enrollment in the study achieved satisfactory glycemic control in the second part of pregnancy due to intensification of treatment. This might explain the lack of capacity of first trimester HbA1c for the prediction of LGA. However, we found that lower levels of HDL and higher levels of triglycerides in the first trimester were both associated with LGA. Importantly, these effects were sustained during the whole pregnancy for HDL and before delivery for triglycerides. The association between decreased HDL and macrosomia in the first half of pregnancy (17–19 weeks) has been demonstrated previously by Clausen et al. in a large cohort of non-diabetic women. However, high triglyceride level was not a risk factor for macrosomia in that study [28]. In our study, we did not find an association between triglycerides and LGA between 20 and 24 weeks of gestation. However, an association was present in the first trimester and before delivery, and requires further study. Unlike Clausen et al., who used non-fasting samples, our women were fasting at the time of blood collection. It eliminated the influence of prandial status on serum triglycerides. However, we cannot exclude the possibility that maternal triglyceride level in mid-pregnancy was altered by rapid improvement of glycemic control related to intensification of treatment that took place between the first and second hospitalizations.

The mechanism by which low HDL might be linked to abnormal fetal growth has not yet been studied. The fundamental function of HDL is to transport cholesterol and other lipids from peripheral tissues to the liver for further processing and excretion [29]. The inverse correlation between HDL and triglycerides has been well documented [30]. In our study, such a correlation was observed throughout the whole pregnancy. Because triglycerides are the main source of non-esterified fatty acids (NEFA) for placental transfer, it can be hypothesized that higher levels of maternal HDL might prevent the fetus from the excessive load of NEFA [31].

Emerging data suggest that many perinatal morbidities are rooted in early pregnancy [32, 33]. Therefore, early prediction of pregnancy complications has become a challenge in obstetrics. It has been shown by Poon et al. that macrosomia can be predicted in the first trimester by a combination of maternal characteristics and laboratory parameters such as PAPP-A and β-hCG used in screening for aneuploidy [34]. However, predictive capacity of this test was weak. Searching for new determinants of LGA, especially in high-risk populations, may help to improve screening performance and, therefore, implement new preventive strategies.

Despite the relatively small sample size, which is the main limitation of our study, the results might be considered population based. This is due to the fact that our department is the only perinatal center for diabetic women in the region and all of them are referred there once pregnancy is confirmed. Nonetheless, we excluded some women due to preeclampsia and some of them were lost to follow-up, mainly due to delivery in another hospital. Thus the incidence of LGA in our study group might not fully reflect its true values in this population. Our decision to exclude women with preeclampsia from the analyzed group was based on the fact that this disease is known to be associated with dyslipidemia [35, 36]. In addition, as shown in other studies, preeclampsia significantly restricts fetal growth potential [37]. Therefore inclusion of those patients could potentially lead to bias in the analysis of the association between maternal lipids and macrosomia.

The strength of the current study is that all women were treated in one center according to the same protocols. Furthermore, all laboratory measurements were performed in the same certified laboratory, thus limiting the possibility of non-random errors resulting from procedural differences.

In conclusion, our study confirms the important role of maternal HDL and triglycerides in promoting excessive fetal growth in women with type 1 diabetes. However, large epidemiological studies are needed to establish outcome-based reference thresholds for lipids in pregnancy. Data derived from such studies would provide the basis for interventional trials aimed at modification of lipids during pregnancy, especially in high-risk patients.