Characteristics of Stillbirths Associated With Diabetes in a Diverse U.S. Cohort

Abstract

Objective

To characterize stillbirths associated with pre–gestational diabetes mellitus and gestational diabetes mellitus (GDM) in a large, prospective U.S. case-control study.

Methods

A secondary analysis of stillbirths enrolled in a prospective, multisite, geographically, racially and ethnically diverse, case-control study in the U.S was performed. Singleton gestations with complete information regarding diabetes status and with a complete postmortem evaluation were included. A standard evaluation protocol for stillbirth cases included postmortem evaluation, placental pathology, clinical testing as performed at the discretion of the health care professional, and a recommended panel of tests. A potential cause of death was assigned to stillbirth cases using a standardized classification tool. Demographic and delivery characteristics among women with pre-GDM and GDM were compared to characteristics of women with no diabetes in pairwise comparisons using chi-square or a two-sample t-test as appropriate. Sensitivity analysis was performed excluding pregnancies with genetic conditions or major fetal malformations.

Results

Of 455 stillbirth cases who were included in the primary analysis, women with stillbirth and diabetes were more likely to be older than 35 years and have a higher body mass index. They were also more likely to have a gestational hypertensive disorder than women without diabetes (28% versus 9.1%; p<0.001). Women with pre-GDM had more large-for-gestational age infants (26% versus 3.4%; p<0.001). Stillbirths occurred more often at term in both women with pre-GDM (36%) and GDM (52%). Maternal medical complications, including pre-GDM and others, were more often identified as a probable or possible cause of death among stillbirths with maternal diabetes (43% vs 4%, p<0.001) as compared to stillbirths without diabetes.

Conclusions

Compared to stillbirths in women with no diabetes, stillbirths among women with pre-GDM and GDM occur later in pregnancy and are associated with hypertensive disorders of pregnancy, maternal medical complications, and large-for-gestational age infants.

Précis

Stillbirths in pregnancies affected by diabetes differ from other stillbirth cases by occurring later in gestation and having greater association with gestational hypertensive disorders and large-for-gestational-age infants.

Introduction

Stillbirth, defined as fetal death at 20 weeks of gestation or greater, occurs in approximately 1 in 160 pregnancies in the United States.¹ Multiple risk factors for stillbirth have been described, including genetic and structural abnormalities, placental abnormalities, infections, and a multitude of maternal conditions, including hypertension, systemic lupus erythematosus, and antiphospholipid syndrome.^2-5

Pre–gestational diabetes mellitus (GDM) is another maternal disease with well-known increased risk for stillbirth.^3,5 While the overall risk for stillbirth in women with pre-GDM has decreased substantially over the past 100 years, about 4% of stillbirths remain attributable to pre-GDM and women with pre-GDM continue to be at increased risk of stillbirth.^6,7 The risk for stillbirth in women with pre-GDM increases with increasing blood glucose and hemoglobin A1C concentrations.^6,8

Major anomalies, placental vascular pathology and changes in fetal metabolic pathways resulting from maternal hyperglycemia are considered as causes of stillbirths in pregnancies complicated by diabetes.⁸ However, a substantial proportion of stillbirths in women with pre-GDM occur without a well-defined reason.^7–9 Our purpose was to characterize stillbirths associated with pre-GDM and gestational diabetes (GDM) in a large, prospective, case-control study as compared to stillbirths in pregnancies without diabetes.

Methods

The Stillbirth Collaborative Research Network (SCRN) conducted a multisite, prospective, case-control study that attempted to enroll all stillbirths and a representative sample of live births at the time of delivery occurring to residents of pre-defined geographic catchment areas. Study conducted at 59 hospitals in portions of five states: Rhode Island, Massachusetts, Georgia, Texas, and Utah. The study enrolled women delivering at 59 hospitals associated within five academic centers and has been described in detail elsewhere.¹⁰ Data collection included maternal interview, medical chart abstraction, and biological specimens. Participants were enrolled between March 2006 and September 2008. All participants provided written informed consent, and approval was obtained from the institutional review boards of each clinical site and the data coordinating center.

For this secondary analysis of the SCRN data, only singleton gestations, with complete postmortem evaluation among stillbirths, and complete information regarding diabetes status were included; only stillbirths were included in the primary analysis. Information regarding diabetes status was considered complete if the patient had been classified as either GDM, pre-GDM or no diabetes. The diagnosis of diabetes in the study population was assessed by chart abstraction. Pre-GDM was determined prior to the initiation of recruitment to be a critical variable for abstraction, and specific fields for pre-GDM and results of one-hour glucose screening and 3 hour glucose tolerance tests were included in the chart abstraction tool used by study personnel. GDM was diagnosed using Carpenter Coustan criteria.¹¹ Women who were diagnosed with diabetes before pregnancy were defined as having pre-GDM. There was no standard approach to classifying women as having pre-GDM or GDM when first diagnosis was made early in pregnancy. Classification in these cases was at the discretion of the primary clinician and insufficient data were available to make a definitive determination. For the purposes of this report, we did not discriminate between type 1 and type 2 pregestational diabetes.

The SCRN protocol included a placental pathology examination for cases and controls, and, for stillbirths, a postmortem examination.¹⁰ These methods were described in earlier publications.^12,13 Network pathologists developed and adopted standardized placental examination protocols, observational criteria, and data collection forms.^12,13 In addition, stillbirths underwent additional clinical testing at the discretion of the primary clinician. Standard testing was recommended for clinicians by the SCRN investigators and included genetic testing, screening for fetal-maternal hemorrhage, antiphospholipid antibody screen, toxicology screen, syphilis serology, and antibody screen.¹⁰ Biospecimens were also collected for subsequent experimental testing.

The Initial Causes of Fetal Death (INCODE) classification system was used to assign causes of death for stillbirths. Details of the algorithms used have been previously described.¹⁴ Important features of the INCODE system include stratification with regard to certainty (probable cause, possible cause and condition present), and the ability to assign more than one cause of death for each stillbirth. The system is not hierarchical. Causes of death were grouped into broad categories including maternal medical conditions (including diabetes), placental conditions, obstetric complications, fetal genetic or structural abnormalities, infection, umbilical cord abnormalities, hypertensive disorders of pregnancy, and others.¹⁴ An algorithm for estimating gestational age at the time of fetal death incorporating the reliability of the estimated due date, ultrasound data, fetal maceration and foot length was used.¹⁵

Characteristics of women, deliveries and neonates among stillbirths with pre-GDM or GDM, were compared pairwise to women with no diabetes using chi-square or a two-sample t-test as appropriate. A sensitivity analysis was performed excluding pregnancies with genetic conditions or major fetal malformations. P-values from a weighted Wald chi-square are reported for comparisons of categorical measures. P-values less than 0.05 are referred to as statistically significant. A supplemental analysis was performed comparing characteristics of stillbirths to live births according to diabetes status. Data analysis was generated using survey-specific procedures in SAS software version 9.4. All analyses were completed with SCRN analytical weights to account for study design.¹⁰ The analytic weights accounted for sampling design and participation rate among study groups with oversampling of preterm and non-Hispanic Black live births. ¹⁰ Self-reported maternal race-ethnicity was collected in maternal interview or from chart abstraction. Women were given race and ethnicity options to choose from or could designate one or more classifications other than those provided. The dataset utilized here classified women as non-Hispanic white, non-Hispanic Black, Hispanic or an “other” category for all other race-ethnicity classifications.

Results

Overall, 663 women with stillbirths were enrolled in the SCRN study, and 512 had complete evaluation and were included in an in-depth cause of death analysis.⁷ Of these, we excluded deliveries prior to 20 weeks gestational age; multiple gestations and missing diabetes information resulting in 455 stillbirth cases being available for the primary analysis (Figure 1).

Figure 1.

Study population. Flowchart depicting inclusion and exclusion of Stillbirth Collaborative Research Network (SCRN) participants in the current study.

*Items not mutually exclusive.

Characteristics of women with stillbirths, compared by diabetes status are shown in Table 1. Women with stillbirth and diabetes were more likely to be older than 35 years of age and had higher body mass index (BMI) than women with no diabetes and stillbirth. Women with stillbirth and pre-GDM were more likely to be non-Hispanic Black compared to women with live births. Women with GDM and stillbirth were more likely to have higher BMI compared women with live births and GDM. (Appendix 1, available online at http://links.lww.com/xxx)

Table 1.

Characteristics of women with stillbirths by diabetes status, compared to women with no diabetes

Characteristic	Value	All diabetes	P-value	Pregestational diabetes	P-value	Gestational diabetes	P-value	No diabetes
Unweighted	N	46		25		21		409
Weighted	N	49		27		22		400
Maternal age (years)	<35	31 (64)	<.001	18 (66)	0.006	14 (62)	0.003	345 (86.1)
	35+	18 (36)		9 (34)		8 (38)		56 (13.9)
Maternal BMI (kg/m2)	Mean (SE)	32.1 (1.0)	<.001	31.4 (1.3)	<.001	32.8 (1.4)	<.001	26.8 (0.33)
Maternal BMI (kg/m2, categorized)	<25 (normal/ underweight)	5 (10)	<.001	3 (14)	0.003	1 (6)	<.001	195 (49.8)
	25–29.9 (overweight)	14 (31)		7 (30)		7 (32)		98 (25.0)
	30–34 (obese)	15 (32)		7 (26)		8 (38)		47 (11.9)
	35+ (morbidly obese)	13 (27)		8 (30)		5 (24)		52 (13.3)
Race Ethnicity	Non-Hispanic white, Hispanic, other	39 (79)	0.679	19 (71)	0.525	20 (90)	0.184	305 (76.3)
	Black, non-Hispanic	10 (21)		8 (29)		2 (10)		95 (23.7)
Marital status	Single	13 (27)	0.982	9 (36)	0.286	3 (17)	0.247	98 (25.9)
	Cohabitating	11 (24)		8 (32)		3 (15)		95 (25.1)
	Married	23 (48)		8 (32)		14 (69)		186 (49.0)
Maternal education	0–11 (less than secondary)	9 (20)	0.05	6 (25)	0.262	3 (15)	0.104	84 (22.3)
	12 (completed secondary)	21 (45)		11 (41)		10 (49)		103 (27.1)
	13+ (college)	16 (35)		9 (35)		7 (36)		191 (50.6)
Insurance	Any public/private assistance	31 (63)	0.527	18 (68)	0.316	12 (56)	0.887	230 (57.9)
	Commercial/ HMO / Military	18 (37)		9 (32)		10 (44)		168 (42.1)
Family income	Any assistance	21 (45)	0.901	13 (49)	0.602	8 (39)	0.684	164 (43.7)
	Only personal income	26 (55)		13 (51)		13 (61)		211 (56.3)
Smoking (3 months prior to pregnancy)	Did not smoke	39 (83)	0.606	21 (80)	0.99	18 (88)	0.385	305 (80.3)
	Smoked	8 (17)		5 (20)		3 (12)		75 (19.7)

P-values reported vs no diabetes

Reported N and p-values based on weighted data

Self-reported maternal race-ethnicity was collected during maternal interview or chart abstraction if interview data were not available. Women were given race and ethnicity options to choose from or could designate one or more classifications other than those provided. The dataset utilized here classified women as non-Hispanic white, non-Hispanic Black, Hispanic or an “other” category for all other race-ethnicity classifications. Race-ethnicity data are pertinent as stillbirth has been shown to vary by maternal race-ethnicity.³

Table 2 shows clinical characteristics of stillbirths in women with pre-GDM and GDM compared to women with no diabetes. Women with diabetes were more likely to have a gestational hypertensive disorder than those without diabetes (28% compared to 9.1%; p < 0.001). The percentage of small for gestational age (SGA) stillbirths was not different in diabetic and non-diabetic pregnancies (9% versus 19.3%; p=0.09). In contrast, more women with pre-GDM and stillbirth had large-for-gestational age infants (26% versus 3.4%; p<0.001) while women with GDM did not (6% versus 3.4%; p=0.594). Fewer intrapartum stillbirths were observed in the combined pre-GDM and GDM group as compared to stillbirths without diabetes (6% versus 18.5%; p=0.049). This is likely due to the higher proportion of intrapartum stillbirths at previable gestational ages in the setting of preterm labor disorders. Many women may not have been screened for GDM at this point in pregnancy and thus diminished the number of intrapartum stillbirths in the diabetes group.

Table 2.

Clinical characteristics of stillbirth deliveries in women with and without diabetes

Characteristic	Value	All diabetes	P-value	Pregestational diabetes	P-value	Gestational diabetes	P-value	No diabetes
Unweighted	N	46		25		21		409
Weighted	N	49		27		22		400
Stillbirth timing	Intrapartum	3 (6)	0.049	1 (5)	0.108	1 (6)	0.227	74 (18.5)
	Antepartum	46 (94)		26 (95)		21 (94)		326 (81.5)
Gestational age at fetal death (weeks)	Mean (SE)	32.1 (1.05)	<.001	30.9 (1.42)	0.039	33.5 (1.47)	<.001	27.8 (0.36)
Sex of baby	Male	25 (52)	0.911	15 (56)	0.719	10 (46)	0.565	208 (52.5)
	Female	24 (48)		12 (44)		12 (54)		188 (47.5)
Birthweight (g)	Mean (SE)	2245 (220.9)	<.001	2183 (321.0)	0.008	2317 (296.8)	<.001	1316 (56.84)
Birthweight (%ile category)	SGA (<10th)	4 (9)	0.085	2 (11)	0.262	2 (8)	0.177	70 (19.3)
	LGA (>90th)	7 (16)	<.001	6 (26)	<.001	1 (6)	0.594	12 (3.4)
Nulliparous	Yes	18 (37)	0.22	13 (48)	0.917	5 (24)	0.046	188 (46.9)
Preeclampsia/ gestational hypertension	Yes	13 (28)	<.001	8 (32)	<.001	5 (24)	0.03	35 (9.1)
Fetal anomalies	Yes	6 (12)	0.977	3 (12)	0.991	3 (12)	0.973	48 (12.0)

P-values reported vs no diabetes

Reported N and p-values based on weighted data

Gestational age differed significantly across diabetes categories with stillbirths in women with diabetes occurring later in gestation than in women without diabetes (Figure 2, p<0.001). Stillbirths occurred most often at 37 weeks or greater in both women with pre-GDM (36%) and GDM (52%). In contrast, the most common epoch for stillbirth in women without diabetes was between 20 and 24 weeks gestation (35%), although women may not have undergone glucose screening at these earlier gestational ages.

Figure 2.

Gestational age at delivery among stillbirths stratified by diabetes status. Percentage of stillbirths among women in each diabetes category in each gestational age epoch.

Combined probable and possible causes of death based on INCODE, compared by diabetes status are depicted in Table 3. As expected, maternal medical complications between all diabetic and non-diabetic women had the most pronounced and significant between-group differences (43% versus 4%, p<0.001). Maternal medical conditions included GDM, pre-GDM, systemic lupus erythematosus, intrahepatic cholestasis of pregnancy, thyroid disorders, renal diseases, asthma, seizure disorders, substance use, shock due to conditions other than infection, antiphospholipid antibody syndrome, thrombophilias and others suspected to be related to the fetal death. Maternal hypertensive disorders including chronic hypertension and hypertensive disorders of pregnancy were analyzed as a separate cause of death category.¹⁴ Differences among women with pre-GDM compared to non-diabetic women had the greatest magnitude of difference (66% versus 4%; p<0.001). In cases with GDM, maternal medical conditions were also more common as a probable or possible cause of death compared to those with no diabetes, but to a lesser degree (16% versus 4%; p=0.025). All pre-GDM and GDM stillbirth cases with a maternal medical condition listed as a cause of death, had either pre-GDM or GDM as the probable or possible cause of death. An analysis of the additional medical conditions was performed. The only clinically significant comorbidity was one case of antiphospholipid antibody syndrome in a pre-GDM patient. (Appendix 2, available online at http://links.lww.com/xxx) Despite a higher prevalence of hypertensive disorders of pregnancy, hypertensive disorders were not more common as a cause of death in stillbirths with diabetes. (Table 3) Given the high proportion of stillbirths at term in the diabetes population an additional analysis was performed to assess differences in cause of death in preterm versus term stillbirths according to diabetes status. Causes of death were similar among these groups with a higher number of deaths due to maternal medical complications in term and preterm stillbirths in diabetic stillbirths (preterm 48% versus 5%; p<0.001, term 35% versus 2%; p=0.013).

Table 3.

Combined probable and possible causes of death (INCODE) by diabetes diagnosis

Cause of death	All diabetes	P-value	Pregestational diabetes	P-value	Gestational diabetes	P-value	No diabetes
Unweighted N	46		25		21		409
Weighted N	49		27		22		400
Placental disease	13 (27)	0.695	9 (32)	0.408	5 (22)	0.758	99 (25)
Infection	4 (8)	0.237	2 (6)	0.233	2 (10)	0.587	56 (14)
Fetal genetic/ structural	7 (14)	0.852	4 (16)	0.911	3 (12)	0.661	60 (15)
Medical complications	21 (43)	<.001	18 (66)	<.001	4 (16)	0.025	18 (4)
Hypertensive disorders	8 (17)	0.068	4 (16)	0.234	4 (19)	0.124	34 (9)
Cord abnormalities	6 (12)	0.81	1 (5)	0.394	5 (21)	0.179	44 (11)
Obstetric complications	5 (10)	0.028	1 (5)	0.026	4 (17)	0.386	105 (26)
Other condition(s)	1 (2)	0.55	1 (3)	0.998	0	-	13 (3)
Unexplained (No cause or present cause only)	5 (11)	0.006	2 (9)	0.023	3 (12)	0.094	102 (25)

P-values reported vs no diabetes

Reported N and p-values based on weighted data

Results were similar when cases with fetal anomalies and genetic abnormalities (n=54) were excluded (data not shown).

Discussion

Stillbirths in women with diabetes occurred at relatively later gestational ages than in women without diabetes. Over a third of stillbirths among women with diabetes occurred after 37 weeks gestation. Although many stillbirths in diabetic women had diabetes assigned (as a component of maternal medical conditions) as a probable or possible cause of stillbirth, several other causes of death were prevalent among pregnancies in women with diabetes including placental abnormalities and hypertensive disorders.

There have been relatively few detailed descriptions of stillbirths in women with diabetes. There are even fewer that distinguish between pre-GDM and GDM. A recent cohort study of 38 stillbirths in women with diabetes in an Australian hospital noted lethal congenital malformations in almost a quarter of stillbirths.¹⁶ This is considerably higher than the one lethal case noted in our cohort. The differences noted may be due to exclusion of pregnancy terminations in the SCRN study, differences in populations, different methods to classify cause of death, and random variation due to small numbers. Other causes of stillbirth in the Australian series were similar to ours and included placental insufficiency, umbilical cord abnormalities and obstetric factors.¹⁶ An older case series of 93 perinatal deaths (stillbirths and neonatal deaths) found asphyxia to be the most common cause of death (73%) in diabetic pregnancies.¹⁷ The reason for attributing the cause of stillbirth to asphyxia was otherwise unexplained in most of the cases in this study. Asphyxia was not considered a cause of death in our cohort, but was likely a contributing factor in cases due to placental abnormalities, and likely reflect differences in classification systems used to assign causes of stillbirth.¹⁴

Others also have noted that a majority of stillbirths in the setting of diabetes occur relatively late in gestation. Some studies suggest that most stillbirths associated with diabetes occur after 37 weeks gestation.^18,19 However risk may begin to increase at somewhat earlier gestational ages, as a large population-based study in the United Kingdom noted an increased risk for stillbirth after 32 weeks gestation in women with pre-GDM.²⁰ In Scotland, the rate of stillbirth in women with pre-GDM remained unchanged despite increased rates of earlier delivery.²¹ Data from the United Kingdom in 2015 showed improvement in the stillbirth rate in diabetic pregnancies, possibly due to earlier delivery (37+0) and tighter glycemic control.²² Accordingly, further research is warranted to determine the optimal gestational age for delivery in pregnancies in women with diabetes.

Notably, over 50% of stillbirths in women with GDM occurred after 37 weeks gestation. Whether GDM is associated with an increased risk of stillbirth is controversial.^23–25 Our study design for the current analysis does not permit us to answer this question. Most stillbirths in women with GDM were ascribed to causes other than diabetes or medical complications.

The relationship between large for gestational age (LGA) infant, diabetes and stillbirth is well recognized.^7,26–28 Available data yield mixed results with regard to whether LGA in non-diabetic women is associated with stillbirth, but overall suggest that there is an association between LGA and stillbirth.^29–36 This association persists when correcting for gestational age at the time of stillbirth.²⁹ Macrosomia may be a surrogate marker for glycemic control or some other factor associated with stillbirth in women with diabetes. High BMI also is a risk factor for stillbirth in women with diabetes as well as an independent risk factor for stillbirth.^3,5,37–40 There were too few small for gestational age (SGA) fetuses in women with diabetes to comment on the lack of association between SGA, diabetes and stillbirth. SGA has been linked to stillbirth in diabetes in other studies.^29,41,42 Additionally, our data showed a higher number of women age 35 years and greater in the diabetic groups, which is also a recognized risk factor for stillbirth.³

Our study had several limitations. We did not have data regarding the quality of glycemic control and glucose tolerance test data was available for only a small subset of patients, precluding meaningful analyses. Also, there was a lack of information about the potential effects of antenatal surveillance or other obstetric interventions. Further, we relied on the clinician’s diagnosis of pre-GDM versus GDM. We could not definitively distinguish between type 1 and type 2 diabetes in some cases and owing to small numbers, these were grouped for analysis. Despite the SCRN being one of the largest and best characterized stillbirth cohorts there were small numbers of patients in the pre-GDM and GDM groups limiting the generalizability of our findings.

The study also had many strengths. The number of stillbirths was large including a sizable percentage whose mothers had diabetes. Inclusion criteria limited bias and the population was racially, ethnically and geographically diverse, enhancing generalizability. Most importantly, each case underwent a thorough evaluation for potential causes of stillbirth and had causes of death assigned using the rigorous INCODE classification system.¹⁴

In summary, most stillbirths in women with diabetes occurred late in gestation and were associated with large for gestational age infants when compared to stillbirths in women without diabetes. They were additionally associated with hypertensive disorders, although hypertensive disorders were not more common as a potential cause of fetal death. This study suggests that more prospective and interventional studies are needed to understand optimal timing of delivery among women with diabetes in order to prevent these stillbirths.