Diabetic Ketoacidosis in Pregnancy: Clinical Risk Factors, Presentation, and Outcomes

Abstract

Context

Diabetic ketoacidosis (DKA) in pregnancy is an obstetric emergency with risk of maternofetal death.

Objective

This work aimed to evaluate DKA events in pregnant women admitted to our inpatient obstetric service, and to examine associated clinical risk factors, presentation, and pregnancy outcomes.

Methods

A retrospective cohort study was conducted at the Mayo Clinic, Rochester, Minnesota, USA, and included women aged 17 to 45 years who were treated for DKA during pregnancy between January 1, 2004 and December 31, 2021. Main outcome measures included maternal and fetal death along with a broad spectrum of maternal and fetal pregnancy outcomes.

Results

A total of 71 DKA events were identified in 58 pregnancies among 51 women, 48 (82.8%) of whom had type 1 diabetes. There were no maternal deaths, but fetal demise occurred in 10 (17.2%) pregnancies (6 miscarriages and 4 stillbirths). Maternal social stressors were frequently present (n = 30, 51.0%), and glycemic control was suboptimal (median first trimester glycated hemoglobin A_1c = 9.0%). Preeclampsia was diagnosed in 17 (29.3%) pregnancies. Infants born to women with DKA were large for gestational age (n = 16, 33.3%), suffered from neonatal hypoglycemia (n = 29, 60.4%) and required intensive care unit admission (n = 25, 52.1%).

Conclusion

DKA is associated with a high rate of maternofetal morbidity and fetal loss. Prenatal education strategies for women with diabetes mellitus should include a strong focus on DKA prevention, and clinicians and patients should have a high index of suspicion for DKA in all pregnant women who present with symptoms that could be attributed to this condition.

Diabetic ketoacidosis (DKA) develops as a consequence of insulin deficiency and is characterized by hyperglycemia, elevated plasma ketones, and metabolic acidosis (1). Although the mechanisms are not entirely understood, the altered metabolic environment of human pregnancy means that DKA can develop more quickly and in the context of less severe hyperglycemia. First, pregnancy is associated with an increase in insulin-antagonizing hormones such as human placental lactogen, estrogen, and prolactin, which contribute to a more than 50% reduction in insulin sensitivity by the third trimester (2-4). In addition, pregnancy is considered a state of accelerated starvation with enhanced lipolysis and ketone body production even with relatively short periods of fasting (5). Pregnant women also experience increased minute alveolar ventilation leading to respiratory alkalosis with compensatory increased renal excretion of bicarbonate and a lower buffering capacity (6). These features, in combination with a relative or absolute insulin deficiency, predispose to DKA development. Several other factors including expanded plasma volume, fetoplacental demands for glucose, and increased glomerular filtration of glucose may result in lesser degrees of hyperglycemia than would be expected for the degree of metabolic disturbance when DKA occurs during pregnancy (1, 7).

Although DKA in pregnancy is considered an obstetrical emergency, robust epidemiological data are limited with reported incidence ranging from 0.5% to 10% of pregnancies affected by pregestational diabetes (8). While associated maternal mortality is rare, reported fetal mortality ranges from 10% to 35% (2, 8). This is illustrated in a recent case-control study from the United Kingdom that identified 82 women with DKA in pregnancy. There were no maternal deaths, but perinatal mortality was 16.0% (1). The largest published cohort in the United States included 64 pregnancies with DKA and reported a similar fetal mortality rate of 15.6%, but maternal outcomes were not examined (4).

The aim of this study is to evaluate DKA events in pregnant women admitted to our inpatient obstetric service at Mayo Clinic, and examine associated clinical risk factors, presentation, and pregnancy outcomes.

Materials and Methods

This was a retrospective cohort study. Following approval from the Mayo Clinic Institutional Review Board, we conducted a search of the Mayo Clinic electronic medical record system using Mayo Data Explorer, a Mayo Clinic–developed data exploration and retrieval tool that facilitates searching of data from multiple clinical and hospital source systems within Mayo Clinic. A major source of data is Mayo Clinic’s Unified Data Platform data warehouse environment. The Unified Data Platform was created from multiple source systems to provide researchers access to Mayo Clinic–wide clinical data in a single centralized database. These source systems include more than 30 years of electronic and scanned medical records including patient demographics, diagnoses, hospital and outpatient clinical notes, and laboratory data. We conducted an initial broad search using the following criteria: all diagnostic codes containing “keto” OR “acidosis” AND all diagnostic codes containing “pregnancy” AND Gender equal to “Female”, “Unknown” or “other” AND “age ≥ 18 years” AND “age ≤ 40 years.” Patients without research authorization and federal medical center patients (prisoners) were excluded. This generated a list of 848 unique individuals who were then screened by manual chart review. We then identified females between the ages 18 and 45 years who were managed for DKA during pregnancy from January 1, 2004 to December 31, 2021. Our start date was chosen to reflect when laboratory results became fully electronic and were therefore consistently available in the medical record for abstraction. Once the cohort was assembled, a broad spectrum of demographic information and clinical variables was manually extracted from the electronic medical record for each DKA episode by 2 physicians (S.M and M.D) following a standardized protocol (using definitions outlined below) with a third physician (A.M.E.) cross-checking the abstracted data for accuracy.

Demographic Measurements

To assess socioeconomic status (SES), we applied the HOUsing-based index of Socio-Economic Status (HOUSES) to our cohort. This is a validated individual-level measure that is derived from publicly available real property data for individual housing units. It provides an individual measure of SES represented by a single factor made up of 4 items (number of bedrooms, number of bathrooms, square footage of the unit, and estimated building value of the unit). The initial measures are aggregated into an overall z score for each individual and assigned a quartile based on all dwellers within the county of residence (quartile 1: lowest SES). Conceptually the measure captures wealth and income along with access to social and environmental resources (9, 10). HOUSES has been associated with a broad range of health outcomes known to be inversely associated with SES (11-14). Maternal social stressors were defined as the presence of at least one of the following three types of social stressors during the pregnancy: physical, sexual, or emotional abuse (15). Body mass index (BMI) was recorded as the BMI at the time of the first obstetric visit for the corresponding pregnancy, or the prepregnancy BMI (within 3 months of pregnancy confirmation), if available.

Outcome Variables

Outcomes were selected based on review of prior studies in this area and a relevant core outcome set (1, 4, 16-18). Gestational hypertension was defined as new onset of hypertension (systolic blood pressure ≥ 140 mm Hg and/or diastolic blood pressure ≥ 90 mm Hg) at 20 weeks or more of gestation in the absence of proteinuria or new signs of end-organ dysfunction (19). Preeclampsia was defined as the new onset of hypertension and proteinuria or the new onset of hypertension and significant end-organ dysfunction with or without proteinuria after 20 weeks of gestation (19). Miscarriage was defined as pregnancy loss before 20 weeks’ gestation and stillbirth was defined as delivery of a fetus at 20 weeks or more of gestation with no signs of life. Neonates were defined as small or large for gestational age (LGA) when their birthweight was below the 10th or above the 90th centile according to the 2017 US reference for singleton birth weight percentiles using obstetric estimates of gestation and adjusted for gestational age of birth and sex (20). The diagnosis of shoulder dystocia was based on the clinician’s assessment in the delivery note documented in the medical record, and neonatal hypoglycemia was diagnosed if the neonatal glucose was less than or equal to 2.5 mmol/L (45 mg/dL). Cases were defined as euglycemic DKA if the maximum recorded venous glucose concentration was less than 13.9 mmol/L (250 mg/dL) (1).

Data were analyzed using BlueSky Statistics (commercial server edition, version 7.40). Comparisons between the pregnancies that resulted in a live birth vs those that ended with fetal demise were evaluated using the chi-square test or Fisher exact test (as appropriate) for categorical variables, and the Wilcoxon Mann-Whitney test for continuous variables (as the relevant variables were not normally distributed based on histograms and Shapiro-Wilk testing).

Results

Maternal Characteristics

A total of 71 DKA events were encountered in 58 pregnancies in 51 women. Table 1 describes maternal characteristics for all pregnancies complicated by DKA and categorizes the pregnancies according to birth outcome (live birth vs fetal demise). The median maternal age for all pregnancies was 28.0 years, and the maternal race was predominantly White (n = 47, 81%). Type 1 diabetes mellitus was present in 48 (82.8%) of all affected pregnancies and type 2 diabetes was present in the remainder. In 7 (12.1%) cases, maternal diabetes was newly diagnosed during the DKA episode. The overall median duration of diabetes was 12.7 years, and the median BMI was 27.7. With the available data, it was possible to measure SES in 70.6% of our cohort (36/51 individuals). The majority (n = 20, 55.6%) were in quartile 1 (lowest relative SES); 8 (22.2%) were in quartile 2; 4 (11.1%) were in quartile 3; and 5 (13.9%) were in quartile 4 (highest relative SES). Maternal social stressors were present in more than half of the pregnancies (n = 30, 51.7%), maternal smoking was reported in 14 pregnancies (24.1%), alcohol use in 2 (3.4%), and drug use in 4 (6.9%). Preexisting diabetes-related complications were relatively rare (retinopathy n = 9 [15.5%], nephropathy n = 6 [10.3%], neuropathy n = 14 [24.1%], and preexisting hypertension 10 [17.2%]), but maternal glycemic control was suboptimal throughout all trimesters of pregnancy (see Table 1). There were no substantial differences in maternal characteristics between pregnancies with live births compared to those with fetal demise.

Table 1.

Maternal characteristics

	All pregnancies N = 58	Pregnancies with live births N = 48	Pregnancies with fetal demise N = 10	P (live births vs fetal demise)
Total number of DKA events	71	58	13
Pregnancies with > 1 DKA event	10 (17.2)	7 (14.6)	3 (30)	.35
Age, y	28.0 (22.3-33.0)	28.0 (22.0-34.0)	27.5 (24.0-30.0)	.67
Duration of diabetes, y	12.7 (3.7-16.9)	13.3 (4.3-17.0)	10.4 (0.9-15.3)	.42
Race
White	47 (81)	41 (85.4)	6 (60)
Other	11 (19)	7 (14.6)	4 (40)	.08
BMI	27.7 (25.27-31.71)	27.6 (24.6-31.1)	28.0 (26.5-33.9)	.38
Gravida	3.0 (2.0-4.0)	3.0 (2.0-4.0)	3.0 (2.0-4.0)	.62
Parity	1.0 (0.0-2.0)	1.0 (0.0-2.0)	1.0 (0.0-1.0)	.32
Use of prepregnancy folic acid	47 (81.0)	40 (83.3)	7 (70)	.38
Social stressors	30 (51.7)	23 (47.9)	7 (70)	.30
Diabetes type
1	48 (82.8)	41 (85.4)	7 (70)
2	10 (17.2)	7 (14.6)	3 (30)	.35
Treatment at DKA diagnosis
Multiple daily injections	34 (58.6)	29 (60.4)	5 (50)	.68
Insulin pump	17 (29.3)	14 (29.2)	3 (30)
None (new diagnosis)	7 (12.1)	5 (10.4)	2 (20)
Smoking during pregnancy	14 (24.1)	11 (22.9)	3 (30)	.69
Alcohol use during pregnancy	2 (3.4)	1 (2.1)	1 (10)	.32
Drug use during pregnancy	4 (6.9)	3 (6.2)	1 (10)	.54
Complications
Retinopathy	9 (15.5)	7 (14.6)	2 (20)	.65
Nephropathy	6 (10.3)	5 (10.4)	1 (10)	≥ .999
Neuropathy	14 (24.1)	11 (22.9)	3 (30)	0.69
Preexisting hypertension	10 (17.2)	8 (16.7)	2 (20)	≥ .999
Maternal glycemic control
Prepregnancy A1c (%)	9.0 (8.1-10.8)	9.0 (7.9-10.8)	9.5 (8.8-9.8)	.67
1st trimester A1c (%)	9.4 (7.4-11.1)	9.3 (7.4-11.1)	9.6 (8.1-10.8)	.65
2nd trimester A1c (%)	7.5 (6.4-8.55)	7.7 (6.6-8.7)	7.1 (6.8-7.9)	N/A
3rd trimester A1c (%)	7.6 (6.5-9.4)	8.1 (6.9-9.4)	9.8 (8.5-11.2)	N/A

Data presented as median (quartile 1-quartile 3) or n (%).

Abbreviations: BMI, body mass index; DKA, diabetic ketoacidosis; N/A, not applicable (inadequate number of data points given 5 out of 10 losses in the fetal demise group occurring in the first trimester).

Characteristics of Diabetic Ketoacidosis Event

Table 2 outlines the characteristics of the DKA presentation for all events according to birth outcome. The median gestational age at time of presentation with DKA was 30.0 weeks, 30.2 weeks for those pregnancies that resulted in a live birth vs 10.9 weeks for those with fetal demise (P = .001). Contributing factors varied between pregnancies with and without fetal demise, with nonadherence most common in those pregnancies resulting in a live birth (44.8% vs 15.4%) and gastrointestinal symptoms more frequently recorded in pregnancies that resulted in fetal demise (30.8% vs 3.4%). Maternal infection was diagnosed in 16 (22.5%) cases, one of which was symptomatic COVID-19 that was diagnosed at 16 weeks’ gestation with subsequent recovery and uncomplicated, spontaneous, normal vaginal delivery at term. Five (7.0%) DKA episodes occurred in the context of glucocorticoid administration in the setting of likely preterm birth, and device failure was reported in 5 (7.0%) cases. Maternal intensive care unit admission was required in 28 (39.4%) women during the DKA episode. There were no substantial differences in key biochemistry results between those episodes associated with a live birth vs fetal demise. In total, 15 (21.1%) episodes were categorized as euglycemic DKA, and none of these episodes were associated with fetal loss. Aside from a lower median glucose (11.3 mmol/L vs 21.3 mmol/L [203 mg/dL vs 384 mg/dL]; P < .001), there were no differences in biochemical parameters between those with euglycemic DKA and those who experienced fetal demise.

Table 2.

Characteristics of diabetic ketoacidosis events

	Total DKA episodes N = 71	DKA episodes in pregnancies with live births N = 58	DKA episodes in pregnancies with fetal demise N = 13	P (live births vs fetal demise)
Gestational age at DKA, wk	30.0 (18.4-33.0)	30.2 (24.6-33.2)	10.9 (6.0-26.7)	.001
Contributing factorsa
Infection	16 (22.5)	14 (24.1)	2 (15.4)
Nonadherence	28 (39.4)	26 (44.8)	2 (15.4)
New diabetes diagnosis	7 (9.8)	5 (8.6)	2 (15.4)
Hyperemesis	6 (8.4)	2 (3.4)	4 (30.8)	.004b
New pregnancy diagnosis	2 (2.8)	2 (3.4)	0
Steroid administration	5 (7.0)	5 (8.6)	0
Device failure	5 (7.0)	5 (8.6)	0
Preeclampsia	1 (1.4)	1 (1.7)	0
Drug abuse immediately before presentation	1 (1.4)	0	1 (7.7)
Severe migraine	1 (1.4)	1 (1.7)	0
Other	5 (7.0)	2 (3.4)	3 (23)
Intensive care unit admission	28 (39.4)	22 (37.9)	6 (46.1)	.82
Maximum anion gap, mEq/L	19.0 (18.0-23.5)	20.5 (18.0-24.0)	19.0 (19.0-21.0)	.85
Maximum BUN, mmol/L mg/dL	4.3 (3.6-5.2) 12.0 (10.0-14.5)	4.4 (3.6-5.3) 12.0 (10.0-14.8)	3.9 (3.2-5.0) 11.0 (9.0-14.0)	.38
Maximum creatinine, µmol/L (mg/dL)	61.9 (53.0-77.8) 0.7 (0.6-0.88)	61.9 (53.0-70.7) 0.7 (0.6-0.8)	61.9 (53.0-97.2) 0.7 (0.6-1.1)	.85
Maximum glucose, mmol/L mg/dL	18.5 (14.4-23.5) 333.0 (259.0-423.5)	18.1 (12.9-23.4) 326.0 (231.3-421.0)	21.3 (17.1-23.8) 384.0 (307.0-428.0)	.15
Minimum bicarbonate, mmol/L	14 (11-16)	13.0 (11.0-16.0)	15.5 (11.8-17.3)	.59
Maximum BOHB, mmol/L	4.1 (2.9-5.2)c	4.1 (3.0-5.3)	4.1 (3.1-4.4)	.81
Maximum osmolality, mean (SD)	291.0 (285.5-298.0)	291.0 (283.3-298.0)	295.0 (289.0-297.0)	.19

Data presented as median (quartile 1-quartile 3) or n (%).

Abbreviations: BOHB, β-hydroxybutyrate; BUN, serum urea nitrogen; DKA, diabetic ketoacidosis.

^a More than one contributing factor in several cases.

^b Considering the 4 most frequent contributing factors.

^c Data missing in n = 13 (18%) cases.

Pregnancy Outcomes

Table 3 outlines the maternal and fetal outcomes for all pregnancies and compares outcomes for live births vs pregnancies with fetal demise. In total, 19 (32.7%) women had gestational hypertension and 17 (29.3%) were diagnosed with preeclampsia. The median gestational age at delivery was 35.5 weeks for all pregnancies (36.7 weeks for pregnancies with live birth and 15.0 weeks for pregnancies with fetal demise; P < .001). Fortunately, no maternal mortality was recorded. Fetal demise occurred in n = 10 (17.2%) of pregnancies (6 miscarriages and 4 stillbirths), and in all cases, the fetal loss was documented at the time of admission for DKA, or within 1 week of dismissal from hospital following a DKA episode (noting that 3 women had ≥ 1 DKA episode during the index pregnancy). The stillbirths occurred at 25 weeks, 28 weeks, 30 weeks and 2 days, and 35 weeks and 3 days gestation. In pregnancies that resulted in live- or stillbirth, induction of labor was performed in 22 (42.3%) pregnancies, while 11 (21.1%) underwent planned, elective cesarean delivery and n = 21 (40.4%) underwent emergency cesarean delivery. Among the pregnancies with live births, 16 (33.3%) neonates were LGA, 4 (8.3%) experienced shoulder dystocia, and 5 (10.4%) were diagnosed with a congenital anomaly. In addition, the median 5-minute Apgar score was 8.0, 29 (60.4%) were diagnosed with neonatal hypoglycemia, and more than half (n = 25, 52.1%) of the neonates required intensive care unit admission with a median duration of stay of 13.7 days.

Table 3.

Pregnancy outcomes

	All pregnancies N = 58	Pregnancies with live births N = 48	Pregnancies with fetal demise N = 10	P (live births vs fetal demise)
Gestational age at delivery, wk	35.5 (33.1-37.18)	36.7 (34.1-37.3)	15.0 (8.8-27.3)	< .001
Maternal mortality	0	0	0
Gestational hypertension	19 (32.7)	19 (39.6)	0	.02
Preeclampsia	17 (29.3)	17(35.4)	0	.03
Birth outcome
Live birth	48 (82.8)	48 (100.0)	0
Miscarriage	6 (10.3)	0	6 (60.0)
Stillbirth	4 (6.9)	0	4 (40.0)
The following variables include live- and stillbirths only	N = 52	N = 48	N = 4
Induction of labor	22 (42.3)	18 (37.5)	4 (100)	.03
Mode of delivery
Normal vaginal delivery	20 (38.5)	16 (33.3)	4(100)	.02
Elective C	11 (21.1)	11 (22.9)	0
Emergency C	21 (40.4)	21 (43.8)	0
The following variables include live births only		N = 48
Birth wt, g, mean		3238.6 (691.7)
Birth wt in relation to gestational age
SGA		0
LGA		16 (33.3)
Shoulder dystocia		4 (8.3)
Congenital anomaly		5 (10.4)
NICU admission		25 (52.1)
Duration of NICU admission, d		13.65 (17.5)
Apgar score at 5 min		8.0 (7.0-9.0)
Neonatal hypoglycemia		29 (60.4)
Neonatal hypoglycemia requiring intravenous glucose		20 (41.7)

Data presented as median (quartile 1-quartile 3) or n (%).

Abbreviations: C, cesarean; LGA, large for gestational age; NICU, neonatal intensive care unit; SGA, small for gestational age.

Discussion

In this study we provide a comprehensive description of 71 DKA events in 58 pregnancies over an 18-year period at one institution. Despite considerable advances in the tools and resources available to manage diabetes mellitus (21, 22), our data reveal that DKA continues to occur during pregnancy and is frequently associated with fetal complications, including pregnancy loss. Our birth outcomes are consistent with prior reports (1, 4) and are underpinned by the poor tolerance of the developing fetus to acidosis. The fetus is likely more sensitive to DKA at earlier gestations, noting the gestational age at DKA of 30 weeks among live births compared to 11 weeks in those that resulted in fetal demise (P = .001). However, as presented, there were also 4 pregnancy losses at more than 20 weeks’ gestation. A prior study associated DKA event severity with increased risk of fetal demise (3). Overall we did not observe a substantial difference in median biochemical results between DKA events that resulted in live birth vs fetal demise. However, within the live birth cohort, 15 DKA episodes can be subclassified as euglycemic DKA. Euglycemic DKA is well described in pregnancy with factors such as increased glomerular filtration rate, enhanced maternal and fetal glucose utilization, and increases in plasma volume contributing to its pathophysiology (2). Although it could be speculated that individuals with euglycemic DKA have a lesser metabolic disturbance and that earlier treatment prevented fetal loss, aside from the magnitude of hyperglycemia, we did not observe any other biochemical differences between the euglycemic and fetal loss groups. Therefore, at this time the authors recommend that health care professionals and women alike maintain a high suspicion for DKA during pregnancy and ensure prompt treatment in all cases.

Reassuringly there were no maternal deaths in our cohort, but the median prepregnancy glycated hemoglobin A₁c (HbA_1c) of 9.0% suggests that women were not well prepared for pregnancy. Maternal social stressors were also frequently present (52% pregnancies) and maternal smoking was reported in 24% pregnancies—far higher than the prevalence of 7.2% reported in the general pregnant population (23). A significant proportion of women in our cohort were in the lowest quartile of SES and this should be considered when assessing an individual’s risk for DKA at the initial prenatal visit. Although lack of information on maternal glycemic control during pregnancy is a limitation of prior studies (3, 6), our trimester-specific data support the inference that high perinatal mortality is due to both the DKA itself and a need for improved diabetes management overall (6). We observe that the maternal HbA_1c remained highly elevated in the first trimester of pregnancy (median 9.4%) and likely contributed to the relatively high proportion of infants born with a congenital anomaly (10.4%). Although in this cohort the HbA_1c improved throughout the second and third trimesters of pregnancy, the median value remained above goal (7.5% in trimester 2 and 7.6% in trimester 3), and consequently we observed high rates of LGA (33.3%) and neonatal hypoglycemia (60.4%) among the live births. In addition to the high rates of fetal mortality, we can also conclude that there is substantial morbidity among infants born to women with DKA during pregnancy. More than half of the live-born infants required admission to the neonatal intensive care unit with an extended median duration of admission of approximately 2 weeks. Maternal morbidity was also high in our cohort, with 29.3% diagnosed with preeclampsia, compared to 11.0% reported in a population-based study of women with pregestational diabetes and similar racial composition (24).

Nonadherence to the recommended insulin program was noted in 39.4% of DKA events. This compares to a nonadherence rate of 51.2% noted in a recent study of DKA in individuals with type 1 diabetes who were not pregnant (25). Nonadherence to insulin therapy can be associated with many factors, including diabetes-related knowledge, lack of family support, fear of hypoglycemia, and SES (26). It is possible that early intervention in the outpatient setting with targeted support to access health care resources including education and medication could reduce the nonadherence rates in this population. We also identified a variety of additional factors that may have contributed to each DKA episode. These include reported device (insulin pump or continuous glucose monitor) failure, concurrent infection, and gastrointestinal symptoms. Increased education and support for women on the importance of seeking urgent assistance in the context of illness, adherence to insulin therapy including sick day rules, and how to promptly address device failure may serve to reduce the risk of DKA in these contexts. However, of perhaps even greater concern, is that corticosteroid therapy (administered for the purposes of fetal lung maturation before anticipated preterm birth) was identified as a precipitant in several cases. Several prior publications have also identified glucocorticoids as a contributor (1, 27), and therefore it is important that health care professionals and patients be aware of the need to substantially adjust insulin doses in the setting of steroid therapy. Individual institutions should adapt published recommendations (28) to develop clinical practice guidelines tailored to their own practice.

Many prior studies of DKA in pregnancy are case reports or case series or lack detailed information on maternal variables (1, 4, 17, 27, 29-33). Although our retrospective study design limits our ability to infer causality, our relatively large cohort contains detailed patient-level data obtained through careful abstraction from an electronic medical record. In this respect our data set is very comprehensive, with less than 5% missing data across the spectrum of end points collected. The variable “maximum β-hydroxybutyrate” is an exception and was missing in 18% of cases (see Table 2) because this test was not routinely used in our clinical practice in the initial years of the study period. It was possible to measure SES in 70% of our cohort, but in combination with our detail on social stressors, we believe that we provide an accurate reflection of the included population. An additional study limitation is that our obstetric practice serves predominately our local population, and in keeping with the background demographics, our cohort primarily contains White women. While this may limit the generalizability of our findings, it is likely that many of the conclusions are relevant to all women with diabetes in pregnancy.

In keeping with prior approaches (1), we included individuals who were treated for DKA during pregnancy. In accordance with our institutional protocol, all women in our cohort therefore received intravenous fluids, insulin, and potassium. Mayo Clinic guidelines recommend the following criteria for diagnosing DKA in pregnancy: serum bicarbonate less than or equal to 15 mEq/dL, anion gap greater than 12 mEq, and serum β-hydroxybutyrate greater than or equal to 3.0 mEq. A glycemic cutoff is not recommended for diagnosis because of the risk of euglycemic ketoacidosis. While we did not use these biochemical characteristics to define entry into our study cohort, a post hoc analysis confirms that 75% of the included cases met 2 out of the 3 criteria listed here. This is broadly in keeping with the most recent publication in this field, in which 57% of cases met a strict definition of DKA (1). This finding reflects the lack of use of β-hydroxybutyrate at our institution in the initial years of our study and appropriate clinician discretion when diagnosing DKA, particularly when considering one or more borderline results within an individual clinical context. We did not identify cases of DKA associated with gestational diabetes mellitus (GDM) in our search of the electronic medical record. In those cases where DKA was the first presentation of diabetes, it was clear that the women had either overt type 1 or type 2 diabetes. The only published population-based study of DKA in pregnancy suggests that DKA occurs rarely in GDM, and even in the small group of women identified as having GDM, the authors could not conclude that type 2 diabetes was comprehensively excluded (1). Therefore, while all women diagnosed with GDM should have a postpartum assessment to rule out type 2 diabetes (34), its importance should be particularly emphasized to those who experience DKA in the index pregnancy.

In conclusion, DKA during pregnancy is associated with high rates of pregnancy loss and considerable maternal and neonatal morbidity. Women who present with DKA are frequently from a lower SES with suboptimal glycemic control before and during pregnancy. All women with diabetes should be counseled on the risk and adverse consequences of DKA during pregnancy, and clinicians should promptly evaluate and treat women who present with suggestive symptoms or signs. As DKA is largely preventable, it is hoped that treatment strategies for diabetes will evolve and reduce the frequency of this condition in the future.

Abbreviations

BMI

body mass index

DKA

diabetic ketoacidosis

GDM

gestational diabetes mellitus

HbA_1c

glycated hemoglobin A_1c

HOUSES

HOUsing-based index of Socio-Economic Status

LGA

large for gestational age

SES

socioeconomic status

SGA

small for gestational age

Financial Support

A.M.E. is supported by the National Institutes of Health (awards DK092721 and HD065987) and the Robert and Elizabeth Strickland Career Development Award in Endocrinology, Metabolism, Diabetes and Nutrition. A.V. is supported by the National Institutes of Health ( awards DK78646, DK116231 and DK126206).

Disclosures

The authors have nothing to disclose.

Data Availability

Some or all data sets generated during and/or analyzed during the present study are not publicly available but are available from the corresponding author on reasonable request.