Continuous Glucose Monitoring and Maternal and Neonatal Morbidity in Pregnant People With Type 1 Diabetes

Abstract

Introduction:

Prior studies have not identified if continuous glucose monitoring (CGM) metrics at a critical gestational age window can discriminate risk of adverse pregnancy outcomes. We evaluated late second- and third-trimester CGM metrics by gestational age associated with pregnancy outcomes in gravidas with type 1 diabetes (T1DM).

Methods:

Dexcom G6 CGM data from a retrospective cohort of singleton gestations with T1DM (2018-2022) at an academic medical center were analyzed. Time in, above, and below range 63 to 140 mg/dL (TIR, TAR, TBR), glycemic variability, and mean glucose concentration were computed in two-week CGM intervals from 24⁰ to 39⁶ weeks^days. Adverse pregnancy outcomes were hypertensive disorders of pregnancy (HDP), large-for-gestational age (LGA), and neonatal hypoglycemia. Linear mixed-effects models were fitted on CGM metrics computed from two-week CGM intervals, with gestational age, adverse pregnancy outcomes (i.e. presence/absence of HDP, LGA, and/or neonatal hypoglycemia), and their interaction as fixed effects.

Results:

In 87 gravidas with preconception median hemoglobin A1c 6.5% (IQR 6.0, 7.1) and maternal body mass index 24.8 kg/m² (IQR 21.9, 27.1), 71% had at least one adverse pregnancy outcome. Between weeks 24⁰ and 37⁶, gravidas with HDP had higher TAR and mean glucose and lower TIR (P < .05). Gravidas with LGA had lower TBR between weeks 24⁰ and 35⁶. TIR, TAR, and mean glucose evolution differed by HDP status, with greatest divergence between groups at 28⁰ to 29⁶ weeks’ gestation (P ≤ .001).

Conclusion:

CGM metrics in the late second to early third trimester, a period of peak insulin resistance, may help to distinguish risk of HDP and LGA in gravidas with T1DM.

Introduction

People with type 1 diabetes mellitus (T1DM) represent a high-risk population who require optimization of glycemic control during pregnancy to mitigate risk of maternal and neonatal morbidity. ¹ The placental hormonal milieu contributes to progressively rising insulin resistance in the second and third trimesters of pregnancy and associated maternal hyperglycemia. ¹ Suboptimal glycemic control in pregnant people with T1DM poses increased risk of hypertensive disorders of pregnancy (HDP), cesarean birth, large-for-gestational age (LGA), preterm birth, neonatal hypoglycemia, and respiratory distress, among other morbidities. ¹ In 2017, the continuous glucose monitoring (CGM) in pregnant women with type 1 diabetes (CONCEPTT) randomized trial demonstrated that the CGM group, compared to the self-monitored blood glucose control group, had increased time within glycemic control target range and reductions in LGA, neonatal hypoglycemia, and neonatal intensive care unit admission. ²

Patients and clinicians have increasingly adopted use of CGM in pregnancy. CGM use helps individuals manage complex insulin dose adjustments, adapt to changes in insulin sensitivity across gestation, respond to day-to-day variability in glucose tolerance and insulin absorption, and mitigate hypoglycemia risk associated with dynamic temporal variability in glucose concentrations, even within a 24-hour period. ² CGM further provides an abundance of data to patients and clinicians regarding their glycemic control within the context of lifestyle factors such as dietary intake, physical activity, sleep, work, and other energy expenditures. ³ Incorporation of CGM data into clinical decision-making can inform day-to-day and hour-to-hour behavioral adjustments as well as insulin dosing, beyond what is typically achieved when using traditional self-monitored blood glucose.

Few studies have evaluated pregnancy outcomes with respect to CGM metrics longitudinally at discrete gestational ages or by trimester of pregnancy and are limited by small sample sizes and in the outcomes assessed.^3-6 The International Consensus on Time in Range has called for more data to demonstrate how CGM metrics correspond to clinical outcomes in pregnancy, as these studies do not identify if a critical gestational age range exists that may distinguish risk of common adverse pregnancy outcomes and/or key windows for tighter glycemic control to mitigate maternal and neonatal morbidity.^1,4,7-9 Therefore, we sought to evaluate the association of the quality of glycemic control, as assessed by CGM metrics, within specific gestational age ranges in the late second and third trimesters and adverse pregnancy outcomes in pregnant people with T1DM.

Methods

This is a retrospective cohort study of a convenience sample of pregnant people with T1DM using the Dexcom G6 CGM system during pregnancy who attended prenatal care in a maternal-fetal medicine clinic at Northwestern Memorial Hospital, an urban, academic quaternary care medical center in Chicago, Illinois, United States. Eligible pregnant people received insulin therapy during pregnancy either via continuous subcutaneous infusion or multiple daily injections, delivered a singleton liveborn neonate after 24⁰ weeks^days between November 1, 2018 and December 31, 2022, were prescribed aspirin 81 mg during pregnancy, and had delivery data, including obstetric and neonatal outcomes, available in the electronic medical record. Individuals in which major congenital anomalies were diagnosed antenatally, who lacked ultrasound confirmation of gestational dating prior to 22⁰ weeks’ gestation, or who did not activate CGM data sharing with obstetric providers through the Dexcom Clarity for Healthcare Professionals online portal were excluded. If individuals had repeat pregnancies during the study period, we utilized data from the earliest pregnancy. The study was approved by Northwestern University’s Institutional Review Board (#STU00217906) before its initiation with a waiver of informed consent.

Data on sociodemographic, clinical, obstetric, and neonatal characteristics were collected through electronic medical record chart abstraction. This additional data included maternal age at delivery, self-reported race and ethnicity (as a proxy for exposure to discrimination), duration of T1DM diagnosis, mode of insulin therapy, history of chronic hypertension or end-organ disease related to T1DM (i.e. retinopathy, gastroparesis, neuropathy, nephropathy), hemoglobin A1c (HbA1c; preconception, antepartum by trimester), and serum creatinine and urine protein:creatinine ratio in the first trimester. For obstetric characteristics, we abstracted parity, pre-pregnancy body mass index (BMI), gestational weight gain, and frequency of estimated fetal weight or abdominal circumference >90^th percentile on last ultrasound prior to delivery. Neonatal characteristics included gestational age at birth, frequency of preterm birth (and indication, as applicable), type of labor onset, mode of delivery, estimate blood loss, and 5-minute Apgar score, neonatal sex and birthweight, and frequency of macrosomia.

Exposures of interest were CGM metrics, including time-above-range (TAR), time-in-range (TIR), time-below-range (TBR), glycemic variability (i.e., CGM coefficient of variation, CV), and mean glucose concentration (mg/dL), assessed in consecutive two-week gestational age intervals (i.e., “bi-weeks”) from 24⁰ to 39⁶ weeks’ gestation. Bi-week intervals were selected as obstetric providers at our center routinely see pregnant people with T1DM and review their glycemic metrics every two weeks throughout the pregnancy. The pregnancy target range for glycemic control was defined as 63 to 140 mg/dL. ⁹

The primary maternal outcome of interest was HDP, including gestational hypertension, preeclampsia with or without severe features (including superimposed preeclampsia), the syndrome of hemolysis, elevated liver enzymes, and low platelet count (HELLP), or eclampsia, using standard definitions per the American College of Obstetricians and Gynecologists. ¹⁰ The primary neonatal outcome was LGA, defined by birthweight ≥90^th percentile using the World Health Organization growth chart for term infants delivered at ≥37⁰ weeks’ gestation or the Fenton growth chart for preterm infants delivered prior to 37⁰ weeks’ gestation.^11,12 The secondary neonatal outcome was neonatal hypoglycemia, considered as a measured glucose level below 40 mg/dL in the first 24 hours of life.

For the overall cohort, we produced frequency distributions for categorical variables and reported the median (interquartile range, IQR) for continuous variables. We used the Chi-square test or Fisher’s exact test for categorical variables, and Wilcoxon rank-sum test for continuous variables, to compare these distributions between individuals who experienced an adverse pregnancy outcome (HDP, LGA, and/or neonatal hypoglycemia) versus those who did not. For all statistical analyses evaluating CGM metrics per bi-week period and adverse pregnancy outcomes, we used available data from those individuals who had bi-week CGM data sufficiency ≥70%, remained pregnant in that bi-week interval (i.e., had not yet delivered), and had maternal and/or neonatal outcome data available. ¹³ We plotted distributions for each CGM metric for each consecutive bi-week period for the overall cohort, as well as stratified by those affected, compared to those unaffected, by each maternal and neonatal outcome. For the HDP outcome, we analyzed only CGM data that preceded the HDP diagnosis.

Linear mixed-effects (LME) models were fitted on each CGM metric (TIR, TAR, TBR, CV, mean glucose concentration), including presence of the considered adverse pregnancy outcomes, gestational age (i.e., each bi-week period), and their interaction as fixed effects, together with an intercept term. The models for all adverse pregnancy outcomes were adjusted for BMI, microvascular complications (i.e. retinopathy and nephropathy), and years from T1DM diagnosis (as a surrogate for T1DM severity). The model for HDP was also adjusted for previous history of HDP, the model for LGA for previous history of LGA, and the model for neonatal hypoglycemia for gestational age at delivery. We included participant-specific random effects for the intercept. The benefit of using LME models is to demonstrate evolutional details of repeated measurements over time (i.e., CGM metrics for each bi-week period), generating an “evolutional curve,” rather than assessing these metrics at one snapshot in time. ¹⁴ The interaction with gestational age (i.e., each bi-week period) provides a valuable assessment, as different groups (i.e., those affected versus unaffected by an adverse pregnancy outcome) could have different evolutional patterns of repeated CGM measures across gestation. From the LME analyses, we report model estimates (standard error) for each CGM outcome (TIR, TAR, TBR, CV, mean glucose concentration) for every bi-week period and in aggregate for the third trimester, comparing the group that experienced versus the group that did not experience the considered adverse pregnancy outcome. Post-hoc analyses based on estimated marginal means were performed to determine the statistical significance of each difference, with a significance level α = 0.05 set for the statistical tests. Adjustment for multiple corrections testing and imputation of missing data were not performed due to the exploratory nature of this analysis.

Data processing, including computation of the CGM metrics, was performed using MATLAB v.R2022b. All statistical analyses were performed using IBM SPSS Statistics v.29.0.1.0. There was no funding source for this study.

Results

Among 162 pregnant people with CGM data available in the Dexcom Clarity online portal during the study period, 92 had T1DM, of whom 87 met inclusion and exclusion criteria. Of these, 57 to 76 pregnant people had CGM data valid for analysis in each late-second/third trimester bi-week period (Supplemental Figure 1).

In this predominately non-Hispanic (90.8%), White (82.8%) cohort of nulliparas (57.5%) with T1DM with median pre-pregnancy BMI 24.8 kg/m² (21.9, 27.1) and median (IQR) preconception HbA1c 6.5% (6.0, 7.1%), 66.7% of pregnant people experienced at least one adverse pregnancy outcomes (HDP, LGA, and/or neonatal hypoglycemia). Baseline sociodemographic characteristics, use of insulin pump therapy (59.8%), and frequency of end-organ disease related to diabetes (17.2%) were similar between individuals who experienced at least one adverse pregnancy outcome and those who did not (Table 1). Individuals with an adverse pregnancy outcome, compared to those without, had higher frequency of preterm birth (25.9% versus 3.4%, P = .01), cesarean birth (63.8% versus 34.5%, P < .01), and macrosomia (32.8% versus 0.0%, P < .001; Supplemental Table 1).

Table 1.

Sociodemographic and Diabetes-Related Clinical Characteristics.

	Overall cohort n (%) or median (IQR) N = 87	No APO n (%) or median (IQR) N = 29	+APO n (%) or median (IQR) N = 58	P-value
Maternal age at delivery, in years	33 (31, 36)	33 (31, 35)	34 (31, 37)	.42
Self-reported race a
White	72 (82.8)	23 (79.3)	49 (84.5)	.55
Not reported	2 (2.3)	-	-	-
Self-reported ethnicity 1
Non-Hispanic	79 (90.8)	28 (96.6)	51 (87.9)	.19
Not reported	2 (2.3)	-	-	-
Duration of type 1 diabetes diagnosis, in years	16 (10, 24)	14.8 (11.7, 17.4)	16.4 (12.2, 20.4)	.03
Mode of insulin therapy b
Multiple daily injections	35 (40.2)	12 (41.4)	23 (39.7)	.88
Chronic hypertension	4 (4.6)	0 (0.0)	4 (6.9)	.12
End-organ disease (at least one of the following)	15 (17.2)	2 (6.9)	13 (22.4)	.08
Retinopathy	11 (12.6)	-	-	-
Nephropathy	2 (2.3)	-	-	-
Gastroparesis	3 (3.4)	-	-	-
Neuropathy	1 (1.1)	-	-	-
Hemoglobin A1c (%)
Preconception (within 3 months of LMP)	6.5 (6.0, 7.1)	6.0 (5.7, 6.8)	6.7 (6.2, 7.2)	.03
1st trimester	6.0 (5.5, 6.6)	5.7 (5.3, 6.1)	6.1 (5.8, 6.7)	<.01
2nd trimester	5.7 (5.3, 6.2)	5.4 (5.1, 5.7)	5.9 (5.5, 6.3)	<.001
3rd trimester	5.9 (5.5, 6.3)	5.6 (5.1, 5.9)	6.1 (5.7, 6.6)	<.001
Postpartum (within 12 months postpartum)	6.2 (5.5, 6.7)	5.6 (5.4, 6.6)	6.4 (5.6, 6.9)	.019
1st trimester urine protein:creatinine ratio	0.02 (0.0, 0.08)	0.0 (0.0, 0.08)	0.06 (0.0, 0.10)	.11
1st trimester serum creatinine, in mg/dL	0.68 (0.61, 0.74)	0.70 (0.62, 0.75)	0.68 (0.61, 0.74)	.64

Abbreviations: APO, adverse pregnancy outcome (at least 1 of: hypertensive disorder of pregnancy, large-for-gestational age, and/or neonatal hypoglycemia); IQR, interquartile range; LMP, last menstrual period.

^aSelf-reported race and ethnicity were abstracted from the electronic medical record, with patient response options as outlined in the table predefined by the electronic medical record software. In the overall cohort, 4 (4.6%) of individuals self-identified as Black and 4 (4.6%) self-identified as Asian.

^bOf the 52 (59.8%) individuals who utilized insulin pump therapy during pregnancy, insulin pump types were identified with the following frequency: 6 traditional (4 Omnipod Eros, 2 Tandem standard); 12 patch (Omnipod DASH), 16 sensor-augmented pump therapy (1 Medtronic MiniMed 530G, 1 Medtronic MiniMed 670G, 15 Tandem t:slim X2 with basal-IQ technology; however, 2 reported turning off the basal-IQ setting and using the t: slim X2 in manual mode throughout pregnancy), 17 Tandem t:slim X2 with control-IQ technology, with hybrid-closed loop capability (6 turned off the control-IQ settings in favor of manual mode throughout pregnancy, and 9 turned off the control-IQ daytime and used manual mode overnight).

Individual adverse pregnancy outcomes occurred with the following frequency: HDP: 33.3%, LGA: 33.3%, and neonatal hypoglycemia: 35.6%. No cases of intrauterine fetal demise, stillbirth, or small-for-gestational age (birthweight <10^th percentile) occurred in this cohort. In the evaluation of glycemic control by bi-weeks across the third trimester, mean TAR and TIR were relatively stable from 24⁰ to 27⁶ weeks’ gestation, then mean TAR decreased and TIR increased with each additional bi-week period after 28⁰ weeks’ gestation. Conversely, TBR and CV decreased from 24⁰ to 27⁶ weeks’ gestation, then remained relatively stable after 28⁰ week’s gestation (Supplemental Figure 2).

In pregnant people that subsequently were assigned a diagnosis of HDP compared to those without HDP, evolution of TAR, TIR, and mean glucose concentration across the late second and third trimester differed in the presence versus absence of HDP. Mean TAR (P < .05) and mean glucose concentration (P ≤ .02) were higher in all bi-week periods from 24⁰ to 37⁶ weeks’ gestation, and mean TIR was lower in most bi-week periods (P ≤ .03; with the exception of weeks 32⁰-33⁶, P = .07), with the greatest divergence for these CGM metrics by HDP status observed at 28⁰ to 29⁶ weeks’ gestation (P ≤ .001, Table 2, Figure 1). Evolution of TBR during this period differed in pregnant people who had HDP compared to those without HDP, with less TBR from 34⁰ to 37⁶ weeks’ gestation (P ≤ .03). Evolution of TBR also differed in pregnant people who had LGA neonates compared to those without LGA, with less TBR from 24⁰ to 35⁶ weeks’ gestation (P < .05, Table 3, Figure 2). While TAR and mean glucose concentration appeared higher, and TIR lower, in pregnant people with versus without LGA neonates, we did not detect a statistically significant difference by LGA status in the evolution of TAR, mean glucose concentration, or TIR across bi-week periods. Pregnant people whose neonates experienced neonatal hypoglycemia had similar evolution of CGM parameters across all bi-week periods as those with euglycemic neonates (Supplemental Table 2, Supplemental Figure 3).

Table 2.

CGM Metrics by Gestational Age and Hypertensive Disorders of Pregnancy Status.

Gestational age in bi-weeks (weeksdays)	TIR (%)		TBR (%)		TAR (%)		CV (%)		Mean Glucose (mg/dL)
	No HDP	HDP	No HDP	HDP	No HDP	HDP	No HDP	HDP	No HDP	HDP
240-256	70.6 (3.2)	62.6 (3.3)	4.2 (0.6)	2.9 (0.7)	25.2 (3.4)	34.5 (3.5)	30.9 (1.3)	32.5 (1.3)	118.0 (4.1)	129.9 (4.1)
	N = 49	N = 20	N = 49	N = 20	N = 49	N = 20	N = 49	N = 20	N = 49	N = 20
	P = .03		P = .09		P = .01		P = .26		P = .009
260-276	71.7 (3.2)	61.9 (3.3)	3.9 (0.6)	2.5 (0.7)	24.4 (3.4)	35.6 (3.5)	30.6 (1.3)	31.9 (1.3)	116.9 (4.1)	131.3 (4.1)
	N = 53	N = 21	N = 53	N = 21	N = 53	N = 21	N = 53	N = 21	N = 53	N = 21
	P = .01		P = .05		P = .003		P = .46		P = .002
280-296	71.2 (3.2)	59.5 (3.3)	3.6 (0.6)	2.5 (0.7)	25.1 (3.4)	37.9 (3.4)	30.0 (1.3)	31.9 (1.3)	117.9 (4.1)	133.9 (4.1)
	N = 52	N = 23	N = 52	N = 23	N = 52	N = 23	N = 52	N = 23	N = 52	N = 23
	P = .001		P = .12		P = .001		P = .17		P < .001
300-316	71.4 (3.2)	62.7 (3.3)	3.6 (0.6)	2.4 (0.7)	25.0 (3.4)	34.8 (3.5)	29.7 (1.3)	30.7 (1.3)	117.7 (4.1)	130.5 (4.1)
	N = 50	N = 21	N = 50	N = 21	N = 50	N = 21	N = 50	N = 21	N = 50	N = 21
	P = .02		P = .10		P = .009		P = .48		P = .005
320-336	72.1 (3.2)	65.6 (3.3)	3.5 (0.6)	2.4 (0.7)	24.4 (3.4)	32.0 (3.5)	29.4 (1.3)	31.5 (1.3)	117.5 (4.1)	128.0 (4.2)
	N = 50	N = 18	N = 50	N = 18	N = 50	N = 18	N = 50	N = 18	N = 50	N = 18
	P = .07		P = .15		P = .047		P = .15		P = .02
340-356	74.3 (3.2)	66.2 (3.4)	3.9 (0.6)	2.2 (0.7)	21.8 (3.4)	31.6 (3.6)	29.6 (1.3)	31.3 (1.4)	114.4 (4.1)	127.4 (4.2)
	N = 53	N = 16	N = 53	N = 16	N = 53	N = 16	N = 53	N = 16	N = 53	N = 16
	P = .03		P = .03		P = .01		P = .25		P = .004
360-376	75.6 (3.2)	66.3 (3.6)	3.6 (0.6)	1.7 (0.8)	20.7 (3.4)	32.0 (3.8)	29.5 (1.3)	30.9 (1.5)	113.1 (4.1)	128.3 (4.5)
	N = 52	N = 10	N = 52	N = 10	N = 52	N = 10	N = 52	N = 10	N = 52	N = 10
	P = .02		P = .02		P = .006		P = .38		P = .002
380-396	77.8 (3.2)	-	3.2 (0.6)	-	19.0 (3.4)	-	29.1 (1.3)	-	112.0 (4.1)	-
	N = 44	N = 0	N = 44	N = 0	N = 44	N = 0	N = 44	N = 0	N=44	N=0
	-		-		-		-		-

Results are reported as the linear mixed-effects model estimates (standard error) for each CGM metric in pregnant people with and without subsequent HDP in each bi-week period, with associated P-values for the pairwise comparison and significant P-values < .05 bolded; N is the number of subjects by HDP status and bi-week period, satisfying sufficient CGM data criteria. Models for HDP were adjusted for pre-pregnancy body mass index, history of HDP, duration of T1DM, and presence of microvascular complications.

Abbreviations: CGM, continuous glucose monitoring; TIR, time-in-range (63-140 mg/dL); TBR, time-below-range (<63 mg/dL); TAR, time-above-range (>140 mg/dL); CV, coefficient of variation (glycemic variability); GMI, glucose management indicator; HDP, hypertensive disorder of pregnancy.

Figure 1.

Evolution of CGM metrics across bi-week periods, stratified by pregnant people with and without a subsequent hypertensive disorder of pregnancy.

The bar plots indicate the mean ± standard deviation of each metric across subjects, for the corresponding bi-week period. A statistically significant difference between the two groups in a specific bi-week is reported with (*) when P < .05 or (**) when P < .005.

Abbreviations: CGM, continuous glucose monitoring; HDP, hypertensive disorder of pregnancy; TIR, time-in-range (63-140 mg/dL); TAR, time-above-range (>140 mg/dL); TBR, time-below-range (<63 mg/dL); CV, coefficient of variation (glycemic variability); Mean, refers to mean glucose concentration.

Table 3.

CGM Metrics by Gestational Age and Neonatal Large-For-Gestational Age Status.

Gestational age in bi-weeks (weeksdays)	TIR (%)		TBR (%)		TAR (%)		CV (%)		Mean Glucose (mg/dL)
	No LGA	LGA	No LGA	LGA	No LGA	LGA	No LGA	LGA	No LGA	LGA
240-256	67.5 (3.5)	62.5 (3.0)	4.7 (0.7)	2.7 (0.6)	27.8 (3.7)	34.9 (3.2)	33.3 (1.4)	31.2 (1.2)	120.9 (4.5)	129.2 (3.9)
	N = 42	N = 28	N = 42	N = 28	N = 42	N = 28	N = 42	N = 28	N = 42	N = 28
	P = .18		P = .009		P = .08		P = .15		P = .08
260-276	68.2 (3.5)	62.7 (3.0)	4.3 (0.7)	2.5 (0.6)	27.6 (3.6)	34.7 (3.2)	32.5 (1.3)	30.2 (1.2)	120.7 (4.4)	128.8 (3.9)
	N = 46	N = 28	N = 46	N = 28	N = 46	N = 28	N = 46	N = 28	N = 46	N = 28
	P = .15		P = .02		P = .07		P = .34		P = .09
280-296	66.0 (3.5)	62.9 (3.0)	4.0 (0.7)	2.4 (0.6)	30.0 (3.6)	34.6 (3.2)	32.5 (1.3)	30.2 (1.2)	123.6 (4.4)	128.6 (3.9)
	N = 48	N = 28	N = 48	N = 28	N = 48	N = 28	N = 48	N = 28	N = 48	N = 28
	P = .41		P = .04		P = .24		P = .12		P = .29
300-316	66.9 (3.5)	64.7 (3.0)	3.9 (0.7)	2.4 (0.6)	29.2 (3.6)	32.9 (3.2)	31.8 (1.3)	29.9 (1.2)	122.2 (4.4)	127.3 (3.9)
	N = 45	N = 28	N = 45	N = 28	N = 45	N = 28	N = 45	N = 28	N = 45	N = 28
	P = .55		P = .04		P = .35		P = .21		P = .29
320-336	67.9 (3.5)	67.5 (3.0)	3.9 (0.7)	2.3 (0.6)	28.1 (3.6)	30.2 (3.2)	31.9 (1.3)	29.6 (1.2)	121.0 (4.4)	125.3 (3.9)
	N = 45	N = 27	N = 45	N = 27	N = 45	N = 27	N = 45	N = 27	N = 45	N = 27
	P = .92		P = .03		P = .60		P = .12		P = .37
340-356	69.9 (3.5)	69.1 (3.0)	4.2 (0.7)	2.6 (0.6)	25.9 (3.6)	28.3 (3.2)	32.0 (1.3)	30.0 (1.2)	118.7 (4.4)	122.6 (3.9)
	N = 48	N = 27	N = 48	N = 27	N = 48	N = 27	N = 48	N = 27	N = 48	N = 27
	P = .83		P = .04		P = .55		P = .17		P = .42
360-376	70.8 (3.5)	70.1 (3.0)	3.7 (0.7)	2.6 (0.6)	25.5 (3.6)	27.2 (3.2)	31.8 (1.3)	29.7 (1.2)	118.1 (4.4)	121.5 (3.9)
	N = 45	N = 27	N = 45	N = 27	N = 45	N = 27	N = 45	N = 27	N = 45	N = 27
	P = .85		P = .17		P = .66		P = .14		P = .47
380-396	73.4 (3.5)	72.8 (3.1)	3.4 (0.7)	2.6 (0.6)	23.1 (3.7)	24.5 (3.3)	30.8 (1.4)	30.0 (1.2)	116.2 (4.5)	118.7 (4.0)
	N = 36	N = 21	N = 36	N = 21	N = 36	N = 21	N = 36	N = 21	N = 36	N = 21
	P = .86		P = .31		P = .73		P = .57		P = .61

Results are reported as the linear mixed-effects model estimates (standard error) for each CGM metric in pregnancies with and without an LGA neonate at each bi-week period, with associated p-values for the pairwise comparison and significant P-values < .05 bolded; N is the number of subjects by LGA status and bi-week period, that satisfied sufficient CGM data criteria. Models for LGA were adjusted for pre-pregnancy body mass index, history of LGA, duration of T1DM, and presence of microvascular complications.

Abbreviations: CGM, continuous glucose monitoring; TIR, time-in-range (63-140 mg/dL); TBR, time-below-range (<63 mg/dL); TAR, time-above-range (>140 mg/dL); CV, coefficient of variation (glycemic variability); LGA, large-for-gestational age.

Figure 2.

Evolution of CGM metrics across bi-week periods, stratified by pregnancies with and without large-for-gestational age neonates.

The bar plots indicate the mean ± standard deviation of each metric across subjects, for the corresponding bi-week period. A statistically significant difference between the two groups in a specific bi-week is reported with (*) when P < .05 or (**) when P < .005.

Abbreviations: CGM, continuous glucose monitoring; LGA, large for gestational age; TIR, time-in-range (63-140 mg/dL); TAR, time-above-range (>140 mg/dL); TBR, time-below-range (<63 mg/dL); CV, coefficient of variation (glycemic variability); Mean, refers to mean glucose concentration.

When assessed in aggregate for the entire period between GA 24⁰ to 39⁶ weeks’ gestation (Supplemental Table 3), mean TAR (P = .002) and mean glucose concentration (P = .001) were higher, while TIR (P = .005) and TBR (P = .04) were lower, in pregnant people with versus those without HDP. Pregnant people with LGA neonates exhibited less TBR (P = .03), but other CGM metrics did not differ between pregnant people with versus without LGA neonates. No difference in mean TAR, TIR, TBR, or mean glucose concentration was identified for pregnant people by neonatal hypoglycemia. No differences in CV were observed between pregnant people with versus without any of the adverse pregnancy outcomes assessed either in bi-week periods or in aggregate after 24⁰ weeks’ gestation.

Discussion

In our single-center cohort, we identified higher mean TAR and mean glucose concentration from 24⁰ to 37⁶ weeks’ gestation, lower mean TIR from 24⁰ to 37⁶ weeks’ gestation, and lower TBR from 34⁰ to 37⁶ weeks’ gestation in pregnancies affected by HDP, compared to those unaffected, with greatest divergence between 28⁰ and 29⁶ weeks’ gestation. Mean TBR was lower from 24⁰ to 35⁶ weeks’ gestation in pregnant people with versus without LGA neonates, with no differences observed in other CGM metrics for this outcome. CGM metrics across the late second and third trimester of pregnancy were similar irrespective of neonatal hypoglycemia status. Differences in CV were not observed for any of the pregnancy outcomes assessed.

Our findings expand upon prior studies evaluating CGM metrics in pregnant people with T1DM. Prior studies have suggested that higher TAR and mean glucose concentration and lower TIR in the second and third trimesters are associated with LGA and neonatal hypoglycemia.^2,5-7,15,16 Meek et al ⁵ also identified higher TAR and mean glucose concentration and lower TIR assessed at 24 weeks’ gestation were associated with preeclampsia, but not when assessed at 34 weeks’ gestation; Sanusi et al ⁷ identified no association of these CGM metrics in pregnancy with preeclampsia. Our findings, both when assessed in aggregate from 24⁰ to 39⁶ weeks’ gestation and across bi-week periods at these gestational ages, stand in contrast to these previously suggested associations.

Disparate findings in our study compared to those in other cohorts may be due to differences in underlying patient co-morbidities, contributions of type of insulin therapy to glycemic control, or other patient-specific characteristics. For the LGA outcome, the small sample size may have limited detection of statistically significant differences in our analysis of these CGM metrics both by bi-week periods and in aggregate. However, as LGA occurs in 30% to 40% of pregnant people with T1DM despite achieving almost normoglycemia, and recognizing that this study population had overall adequate glycemic control (median HbA1c ≤6.0%) across all three trimesters, it is likely factors other than glucose levels are at play in development of LGA.^6,17 For the neonatal hypoglycemia outcome, it is also possible that tight intrapartum glycemic control may have diminished any potential associations with antepartum glycemic control in this cohort of CGM users.

Our analysis more closely aligns with work by Scott et al, ⁴ who evaluated evolution of CGM metrics weekly from early to late pregnancy in 386 pregnant people with T1DM. They identified increased divergence in CGM metrics from 20 to 30 weeks’ gestation, with higher TAR and mean glucose concentration and lower TIR, TBR, and CV in pregnancies with versus without LGA; this divergence between groups persisted but plateaued around 34 to 36 weeks’ gestation. We similarly identified less TBR after 24 weeks’ gestation, though not other CGM metrics, was significantly associated with pregnancies resulting in LGA neonates, and that the ability of CGM metrics to distinguish between LGA and additional adverse pregnancy outcomes dissipated after 35 weeks’ gestation. Our findings, similar to Scott et al, ⁴ reflect physiologic changes in which insulin resistance increases in the late second and early third trimesters of pregnancy, then plateaus by 36 weeks’ gestation.^1,18

Our study has notable strengths and limitations. We demonstrate how CGM metrics differ and evolve over each advancing two-week gestational age interval in the third trimester in pregnant people with T1DM affected by clinically relevant adverse pregnancy outcomes; few prior studies have achieved this degree of discrete temporal analysis of evolution of CGM metrics. Our study is novel in using advanced statistical methods to analyze more than a million raw CGM datapoints to identify critical periods in the late second and early third trimesters where divergence of TIR below the 70% recommended threshold distinguishes those pregnant people and their neonates at risk of common adverse pregnancy outcomes, particularly HDP. ⁹ We are able to analyze clinically useful CGM metrics, accounting for adequate data sufficiency, in a contemporary cohort that reflects current diabetes technology and antenatal management practices, including users of both insulin pump therapy and injections.

However, our statistical power and ability to perform subgroup or sensitivity analyses to account for potentially confounding clinical characteristics (e.g. mode of insulin delivery with multiple daily injections versus insulin pump therapy; hybrid closed-loop versus non-automated insulin pump systems), or other important clinical outcomes (e.g., small-for-gestational age) are limited by the small sample size. Additionally, as a single-site observational study using retrospective data without a control group, our results must be interpreted with caution as we cannot determine causality and can only demonstrate associations between the CGM metrics and pregnancy outcomes assessed. Moreover, our findings are not generalizable to pregnant people with type 2 or gestational diabetes, who may exhibit different patterns of CGM metric evolution that distinguish risk of adverse pregnancy outcomes given differences in the underlying pathophysiology of these diabetes phenotypes. Finally, additional evaluation in more diverse populations with variable sociodemographic backgrounds and degree of health literacy is needed considering concerns surrounding health equity and access to CGM and other diabetes technology in pregnant people in the United States.

Despite these limitations, our study findings enhance the current understanding of how CGM targets at varying gestational ages in late pregnancy, especially during the period of peak insulin resistance in the early third trimester, may differentially portend risk of adverse pregnancy outcomes. Furthermore, as graphically represented, these data provide pregnant people with T1DM and their providers with bi-weekly CGM targets for which to aim throughout the late second and early third trimesters, beyond HbA1c alone, to optimize glycemic control, which may help avert the development of adverse pregnancy outcomes.^1,4,7,8 Overall, our findings support that the trend of bi-weekly CGM targets in pregnant people with T1DM discretely characterizes glycemic control across gestation, and can be used to inform patient counseling regarding risk of adverse pregnancy outcomes and critical windows for tightened glycemic control.