Risk of Neonatal Hypoglycemia in Infants of Mothers With Gestational Glucose Intolerance Free

Hypoglycemia is the most common neonatal metabolic abnormality. Prompt diagnosis and management of neonatal hypoglycemia are necessary to avoid adverse sequelae, most notably long-term neurodevelopmental impairments (1–6). It has been estimated that 51% of infants at risk experience neonatal hypoglycemia in the first 48 h of life. The incidence of neonatal hypoglycemia among infants considered not at risk is largely unknown due to current recommendations to screen only infants at risk (7), such as those born to people with diabetes, either pregestational diabetes or gestational diabetes mellitus (GDM) (2,8).

A two-step approach is commonly used in the U.S. to screen pregnant people for GDM. At 24–28 weeks’ gestation, a 1-h 50-g glucose load test is administered. Blood glucose levels ≥130–140 mg/dL (depending on the institution) 1 h after a glucose load are considered abnormal, and a 3-h 100-g oral glucose tolerance test is performed after an abnormal screening test (9.10). A diagnosis of GDM is made when two or more values on the glucose tolerance test are abnormal. The glucose tolerance test is commonly interpreted based on Carpenter-Coustan criteria (plasma glucose values: fasting ≥95 mg/dL, 1 h ≥180 mg/dL, 2 h ≥155 mg/dL, and 3 h ≥140 mg/dL) (9).

It is estimated that 45% of infants born to people with GDM experience neonatal hypoglycemia (11). Although clinical GDM diagnosis is binary, in studies (most notably the Hyperglycemia and Adverse Pregnancy Outcome [HAPO] study) continuous associations have been reported between maternal glycemia during pregnancy and adverse outcomes, including increased birth weight, risk of cesarean delivery, and neonatal hypoglycemia (12). These findings suggest that gestational glucose intolerance (GGI), or glucose intolerance below the diagnostic threshold for GDM, may also increase risk of neonatal hypoglycemia; however, the current evidence is sparse, mixed, and based on inconsistent study methods (13–16).

Given the importance of identifying infants at risk for neonatal hypoglycemia to inform screening recommendations, the overall goal of this study was to examine the relationship between GGI and neonatal hypoglycemia in two combined cohorts of mother-infant dyads.

This was a secondary analysis of two combined electronic medical record–derived cohorts of pregnant people and their infants with prenatal care and delivery in the Mass General Brigham health system (Boston, MA): the glycated CD59 (GCD59) and the Maternal Health Cohort (MHC) studies. Both studies were approved by the Mass General Brigham Institutional Review Board (GCD59, protocol no. 2011P002254, and MHC, protocol no. 2018P002283).

The GCD59 study included discarded plasma samples and electronic medical record–derived data from pregnant individuals (n = 3,853) undergoing prenatal care at Brigham and Women’s Hospital (BWH) between October 2019 and March 2023. Pregnant individuals were eligible for enrollment if they had consented to make their excess clinical samples and associated clinical data available to the Crimson Biobank, had discarded clinical plasma sample available at <14 weeks’ gestation, and were between 18 and 50 years old (17). Only the first pregnancy per mother was included in the data set.

The MHC is a repository of electronic medical record–derived pregnancy data from individuals who received prenatal care at Massachusetts General Hospital (MGH) between 1998 and 2016 and their infants’ electronic medical records (18–20). Data from deliveries that occurred between 1 January 2014 and 31 December 2015 (n = 7,000) were included in this analysis. Only the first pregnancy per mother was included in the data set.

To align inclusion criteria between cohorts, we limited the study population to individuals between 18 and 50 years old in the MGH cohort (n = 95 [1.4%] excluded). We additionally excluded multiple gestation pregnancies (BWH, n = 134 [3.5%] excluded, and MGH, n = 206 [2.9%] excluded) and those without complete glucose tolerance testing data available (BWH, n = 459 [11.9%] excluded, and MGH, n = 1,697 [24.2%] excluded) (Supplementary Fig. 1). Complete glucose tolerance testing data available included documentation of a glucose load test and, if indicated, a complete glucose tolerance test. Among individuals with incomplete glucose tolerance testing, 1,846 had no glucose load test, 277 had a high glucose load test and no glucose tolerance test, and 33 had an incomplete glucose tolerance test. We excluded individuals with preexisting type 1 and type 2 diabetes by requiring complete glucose tolerance testing data, as people with known preexisting diabetes do not undergo GDM screening (18).

The final cohort included data from 3,260 dyads enrolled in the BWH cohort and 5,002 dyads enrolled in the MGH cohort. Data from each cohort were concatenated resulting in a final analytic sample size of 8,262 mother-infant dyads.

The ascertainment of glucose tolerance status, hypoglycemia outcomes, and/or other data have previously been described for both cohorts (17,18). Clinical data from the prenatal obstetrics visits, delivery hospitalization, and postpartum visits were available for both cohorts.

Our primary exposure was gestational glucose tolerance, categorized as normal glucose tolerance (NGT), GGI, or GDM, derived from the glucose load and tolerance test results. We defined NGT according to a normal glucose load test result per institutional policy (<130 mg/dL at BWH and <140 mg/dL at MGH). GGI was defined according to an elevated glucose load test result (≥130 mg/dL at BWH and ≥140 mg/dL at MGH) and no or one abnormal value on the glucose tolerance test. GDM was defined according to an abnormal glucose load test result plus two or more elevated values on the glucose tolerance test standardized with use of Carpenter-Coustan criteria (plasma glucose values: fasting ≥95 mg/dL, 1 h ≥180 mg/dL, 2 h ≥155 mg/dL, and 3 h ≥140 mg/dL (9). Our secondary exposure was GGI subtype, comparing NGT with GGI-0 (abnormal glucose load test result, no abnormal glucose tolerance test values), GGI-1 (abnormal glucose load test result, one abnormal glucose tolerance test value), and GDM.

The outcome for this analysis was neonatal hypoglycemia defined as any blood glucose level during the delivery hospitalization <45 mg/dL or a documented diagnosis of neonatal hypoglycemia based on ICD-9 or ICD-10 codes. A cutoff of 45 mg/dL was chosen based on the American Academy of Pediatrics–recommended target blood glucose prior to routine feeds and previous research (8,21). ICD-10 codes E16.2, P70.3, and/or P70.4 were used for both cohorts. Additionally, ICD-9 codes 775.6 and/or 215.2 were used for the MGH cohort due to data obtained from prior to the use of ICD-10 codes. Infants with all blood glucose values ≥45 mg/dL or who did not have a blood glucose measurement or ICD-9 or -10 diagnosis were classified as normoglycemic (except for in sensitivity analyses). In instances where all recorded blood glucose values were ≥45 mg/dL but an ICD-9 or -10 diagnosis was present, the infant was considered to be hypoglycemic.

Neonatal hypoglycemia screening criteria were similar across institutions. According to institutional policy at BWH during the study period, testing was indicated in infants who were symptomatic, preterm, born to individuals with GDM or diabetes, small for gestational age or large for gestational age, in respiratory distress for >1 h, born at >42 weeks’ gestation, or born to mothers treated with β-blockers or terbutaline. Additional criteria included a 5-min Apgar score <7, birth weight <2,500 g, and a family history of hypoglycemia, midline abnormalities, or congenital syndrome associated with hypoglycemia. According to institutional policy at MGH during the period of the study, testing was indicated in infants who were symptomatic, were preterm, were born to mothers with GDM or diabetes, were small for gestational age or large for gestational age, had respiratory distress for >1 h, or had a temperature <97.7°F twice despite warming.

Covariates including maternal age, race/ethnicity, insurance status, parity, height, weight, mode of delivery, infant sex, and gestational age at delivery were abstracted from the electronic medical records. We used maternal height and initial prenatal visit weight to calculate maternal BMI (calculated as weight in kilograms divided by the square of height in meters) for BWH participants. For MGH, we used maternal height and weight to calculate BMI and estimated BMI at 12 weeks’ gestation using a previously described algorithm (19). We defined preterm delivery according to delivery occurring before 37 weeks’ gestation. Small for gestational age and large for gestational age were determined with use of the sex-specific growth curves of Oken et al. (22) as a birth weight percentile <10 or >90.

Analyses were performed with SAS 9.4. We assessed data distributions using visual displays of data; assessment of means and SDs, medians and interquartile ranges (IQRs), skewness, and kurtosis; and the Shapiro-Wilk test of normality. We described sociodemographic and clinical characteristics of the cohort using medians (IQRs) for continuous variables and counts and percentages for categorical variables.

We examined associations between gestational glucose tolerance category and neonatal hypoglycemia using logistic regression analysis, unadjusted and adjusted for covariates, including maternal age, parity, and BMI. We chose to adjust for these covariates because they differed across both exposure and outcome categories and were not potentially causal. We performed logistic regression analyses using gestational glucose tolerance category as a three-level variable (NGT, GGI, GDM) and a four-level variable with GGI subtypes (NGT, GGI-0, GGI-1, GDM). For all analyses, NGT was the reference group. We included interaction terms for infant sex and preterm delivery and performed stratified analyses for variables with significant interaction terms.

Because not all infants were screened for hypoglycemia, we performed additional sensitivity analyses restricted to infants with a blood glucose measurement only, a blood glucose measurement and/or ICD diagnosis of neonatal hypoglycemia, and a risk factor for neonatal hypoglycemia, including large-for-gestational-age birth weight, small-for-gestational-age birth weight, preterm delivery, or maternal GDM. To further address potential ascertainment bias, we developed an additional sensitivity analysis, treating the infants who did not have a blood glucose measurement or ICD diagnosis of neonatal hypoglycemia as missing outcome data, addressed through multiple imputation. Using Δ-adjusted multiple imputation (23), we assessed how results changed under varying assumptions about the missing data mechanism, including scenarios where the data are permitted to be missing not at random (Supplementary Material). In these analyses, we considered scenarios where between 1% and 33% of neonates who were unscreened actually had hypoglycemia, with the upper bound of 33% chosen because this was the rate of hypoglycemia among infants with a blood glucose measurement or ICD diagnosis of neonatal hypoglycemia.

For all analyses, the significance threshold was designated a priori as P < 0.05.

The data sets generated during or analyzed in the current study are available from the corresponding author on reasonable request after institutional review board approval and an executed institutional data transfer agreement.

Maternal and infant characteristics across gestational glucose tolerance categories are presented in Table 1. Among the 8,262 dyads included, 82.3% of pregnant individuals had NGT, 13.9% had GGI, and 3.8% had GDM. Maternal age and BMI, as well as the percentage of preterm infants, cesarean deliveries, and large-for-gestational-age infants, increased across glucose tolerance categories.

The frequency of neonatal hypoglycemia in each of the gestational glucose tolerance categories is presented in Fig. 1. Approximately 12% of all infants had neonatal hypoglycemia. The proportion of infants with neonatal hypoglycemia increased across gestational glucose tolerance categories: 10.0% of infants born to mothers with NGT, 13.7% of infants born to mothers with GGI, and 38.1% of infants born to mothers with GDM had neonatal hypoglycemia. The proportion of infants with neonatal hypoglycemia was higher for the GGI-1 (17.5%) than for the GGI-0 (12.6%) subtype.

Associations between maternal glucose tolerance status and neonatal hypoglycemia are presented in Fig. 2 and Supplementary Table 1. With adjustment for maternal age, parity, and BMI, infants born to mothers with GGI had 1.35 times higher odds (95% CI 1.11, 1.64) of neonatal hypoglycemia in comparisons with those born to mothers with NGT. In adjusted analyses with maternal glucose tolerance assessed as a four-level variable, infants born to mothers with GGI-0 had 1.28 times higher odds (95% CI 1.03, 1.60) and GGI-1 1.58 times higher odds (95% CI 1.11, 2.25) of neonatal hypoglycemia in comparisons with those born to mothers with NGT (Fig. 2 and Supplementary Table 1). Infants born to mothers with GDM had 4.89 times higher odds (95% CI 3.81, 6.28) of neonatal hypoglycemia.

We found that gestational age and infant sex were significant effect modifiers (interaction P < 0.05) and conducted stratified analyses in these subgroups (Table 2). The associations for GGI remained significant for term infants only (term odds ratio [OR] 1.40 [95% CI 1.13, 1.73] and preterm OR 1.02 [95% CI 0.61, 1.70]). The association was also only significant for male infants (male infants OR 1.41 [95% CI 1.11, 1.80] and female infants OR 1.22 [95% CI 0.89, 1.69]) (Table 2), possibly related to a higher proportion of male infants with risk factors (24.8% vs. 22.0%) and, thus, being screened (24.8% vs. 22.0%).

The associations with GGI were attenuated toward the null for all three subgroups with analyses limited to infants with a blood glucose measurement only (OR 1.18 [95% CI 0.94, 1.47]), a blood glucose measurement and/or ICD diagnosis of neonatal hypoglycemia (OR 1.12 [95% CI 0.89, 1.39]), and a risk factor for neonatal hypoglycemia (OR 1.08 [95% CI 0.82, 1.44]) (Supplementary Table 2), though the CIs still overlapped with that from the primary analysis.

In treating the infants without blood glucose measurements or ICD diagnoses of neonatal hypoglycemia as having missing outcome data (addressed through multiple imputation), estimates of the OR for GGI ranged from 1.19 (33% of infants with missing data imputed as having neonatal hypoglycemia) to 1.33 (1% of infants with missing data imputed as having neonatal hypoglycemia). Full results can be found in Supplementary Material.

From a secondary analysis of two combined cohorts of mother-infant dyads, we report that infants born to individuals with GGI had higher odds of neonatal hypoglycemia compared with infants born to individuals with NGT. Further, increasing severity of GGI was associated with higher odds of neonatal hypoglycemia. Findings were consistent with those of the primary analysis when we treated infants without a blood glucose measurement or ICD diagnosis as missing outcome data, addressed through multiple imputation, assuming up to 33% of unscreened infants had neonatal hypoglycemia. Our findings support the evaluation of maternal glycemia along a continuum, as opposed to a binary assessment, in considering infant risk for adverse outcomes such as hypoglycemia.

Our findings are in line with findings of previous research indicating that GGI may be an at-risk state for neonatal morbidities, including complications related to large for gestational age and birth weight (20). Compared with previous studies where investigators examined the relationship between subclinical maternal glucose abnormalities and neonatal hypoglycemia, in our study we used a broader categorization of GGI including any pregnant individual with an abnormal glucose load test but normal subsequent glucose tolerance test. Previous studies included more stringent definitions, with categorization of subclinical maternal glucose abnormalities based off individual abnormal test values, such as 1-, 2-, or 3-h results, or without categorization of an abnormal fasting glucose level on the glucose tolerance test as a glucose abnormality (13–16). Wang et al. (13) reported that for infants born to mothers with subclinical glucose abnormalities, defined as an abnormal 1-h value on the oral glucose tolerance test, there was a higher incidence of neonatal hypoglycemia compared with incidence among those born to mothers with NGT. No difference in incidence of neonatal hypoglycemia was found with glucose intolerance defined according to an abnormal 2- or 3-h value or according to abnormal glucose load test levels with normal glucose tolerance test results. Corrado et al. (14) found that the prevalence of neonatal hypoglycemia was similar among infants born to individuals with one or zero abnormal values on the glucose tolerance test. Kim et al. (15) reported that among pregnant individuals with an abnormal glucose load test result, for infants born to individuals with an abnormal 2-h value on the glucose tolerance test there was higher prevalence of neonatal hypoglycemia (blood glucose <35 mg/dL) compared with the prevalence for those with no elevated glucose tolerance test values. No differences were found between those with a 1- or 3-h elevated value. Gluck et al. (16) reported a similar prevalence of neonatal hypoglycemia (blood glucose <40 mg/dL) between infants born to individuals with one abnormal value of the last three glucose tolerance test measurements (excluding abnormal fasting value) and those born to individuals with NGT.

We found that the associations between glucose intolerance, including both GGI and GDM, and neonatal hypoglycemia were significant in term infants only, indicating that glucose intolerance may not be associated with a further increased risk of hypoglycemia in infants born preterm. Similarly, findings were attenuated when analyses were restricted to infants with any risk factor for hypoglycemia. Other factors, such as early provision of dextrose containing fluids, may impact hypoglycemia predisposition or protection in preterm infants, although we could not determine this conclusively in our study. Our finding that GGI increased risk of neonatal hypoglycemia in male infants only is consistent with reports from previous studies that male infants are more likely to develop neonatal hypoglycemia, in both univariate (7,24) and multivariate (25,26) analyses. In addition, male infants were more likely to have a risk factor and be screened for hypoglycemia compared with female infants, which might partly explain this finding.

A strength of our study was that we examined risk of neonatal hypoglycemia by increasing degree of GGI through inclusion of a secondary exposure with two levels of GGI, zero or one abnormal value on the glucose tolerance test. Our sample size was large, which allowed adequate power to adjust for potential confounders and examine associations within important subgroups. Another strength of our study was that our sample was derived from some of the largest delivery centers in MA, which improves the generalizability of our results. We also used the clinical standard of <45 mg/dL for neonatal hypoglycemia diagnosis in the U.S., ensuring results are generalizable to similar hospital settings (8). Lastly, there are little pragmatic data available within the current clinical context with use of available clinical laboratories to assess risk, which this study provides.

Our study did have limitations. As data for both cohorts were abstracted from the electronic medical record, our study relied on accurate and complete clinical documentation, especially for ICD coding. To address this, we performed sensitivity analyses where we restricted our analysis to infants with a blood glucose measurement available. Our study also relied on clinical methods for gestational glucose tolerance screening, and there were differences between sites in the threshold considered abnormal for the glucose load test result. Although these differences suggest that participants from the MGH cohort categorized as having GGI may have slightly higher levels of glucose intolerance than those in the BWH cohort, these differences reflect true variability in screening methods across hospitals in the U.S. At both institutions, a 100-g oral glucose tolerance test and Carpenter-Coustan criteria were used to diagnose GDM, as this is the most common way of diagnosing GDM in the U.S. Since neither institution used a 75-g oral glucose tolerance test or International Association of the Diabetes and Pregnancy Study Groups (IADPSG) criteria (as is more common outside of the U.S.), we cannot comment on how using these cutoffs would impact our findings. However, IADPSG cut points for diagnosis of GDM are lower and generally result in more women being diagnosed with GDM (27,28). Findings from randomized controlled trials of two-step versus one-step GDM screening methods show that the one-step approach results in diagnosis of a higher proportion of infants with hypoglycemia (28–30), likely due to increased ascertainment, since GDM diagnosis is a screening criterion. We also limited our analysis to individuals with complete glucose tolerance testing data, which was necessary to test our hypotheses. This may have resulted in a lower-than-expected GDM rate. For example, it is common practice at BWH to diagnose individuals with GDM without conducting a glucose tolerance test if their glucose load test level is ≥190 mg/dL. In our study, these individuals were categorized as having incomplete glucose tolerance testing and were removed from the analysis, which may have lowered the rate of GDM in our study population. Incomplete glucose screening may also be due to clinical data entry errors, patient preference, and conditions resulting in inability to complete screening, such as nausea or bariatric surgery (19,31). We previously reported within MHC that 25% of GDM cases identified by ICD code could not be confirmed with laboratory values (18), and others have reported that 34% of GDM cases at BWH are diagnosed outside of clinical guidelines (31). Additionally, the national prevalence of GDM increased from 6% in 2016 to 8.3% in 2021 (32); thus, our GDM diagnosis rate in part reflects historical rates of GDM, given that more than one-half of our cohort delivered in 2014–2015. Our study also relied on clinical screening methods for neonatal hypoglycemia where infants were only screened if they had a risk factor or symptoms. As a result, we were unable to capture neonatal hypoglycemia events that may have occurred in unscreened infants, which may have introduced ascertainment bias. To address this, we performed a sensitivity analysis in which we used Δ-adjusted multiple imputation, which permitted the neonatal hypoglycemia data among unscreened infants to be missing not at random. Associations in this sensitivity analysis remained consistent with those observed in the primary analysis (95% CI for GGI OR in the primary analysis, 1.11–1.64; OR in Δ-adjusted multiple imputation sensitivity analysis, 1.19–1.33). Lastly, our study did not include assessment of the impact of these findings on later child health outcomes.

Future studies should be conducted to assess the relationship between GGI and neonatal hypoglycemia in a cohort with universal neonatal blood glucose ascertainment. Leveraging technology such as continuous glucose monitoring for both pregnant people and infants would allow for detailed understanding of glycemic trends. Lastly, in future studies investigators should examine the long-term neurodevelopmental impact of neonatal hypoglycemia in infants born to people with GGI to determine the clinical significance of these findings for long-term child health.

Our study provides evidence that GGI increases risk of neonatal hypoglycemia and suggests that there is a dose-dependent response between increasing degree of GGI and risk of neonatal hypoglycemia, in a large, U.S.-based population. These findings have implications for prenatal and neonatal care practices and should be the focus of future studies in large cohorts with detailed maternal, neonatal, and child characterization.

Acknowledgments. The authors thank Dr. Ryan Simpson (Friedman School of Nutrition Science & Policy, Tufts University, Boston, MA) for helpful discussions and feedback throughout the course of this analysis. Dr. Simpson received no financial support for his participation.

C.E.P. is an editor of Diabetes Care but was not involved in any of the decisions regarding review of the manuscript or its acceptance.

Funding. The GCD59 study was supported by National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, grant R01 DK118528. The MHC study was supported by the MGH Claflin Distinguished Scholar Award and the MGH Physician Scientist Development Award. J.M. was supported by a National Institutes of Health T32 training grant for endocrinology (T32DK007028- 47S1).

Duality of Interest. J.H. is a founder and has financial interest in Mellitus. Mellitus had licensed related intellectual property and has interests in developing diagnostic tools for diabetes. C.E.P. receives payments from Wolters Kluwer for UpToDate chapters on diabetes in pregnancy and has received payments for consulting and speaking from Mediflix. S.S. received devices from Dexcom for a research study at no cost. The interests of J.H. and S.S. were reviewed and are managed by BWH and Brigham Health in accordance with their conflict of interest policies. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.A. and T.T. analyzed data. C.A. wrote the first draft of the manuscript. J.M., C.C.M.S., S.H., T.T., K.E.J., J.H., C.E.P., and S.S. reviewed and edited the manuscript. All authors reviewed the manuscript for accuracy and read and approved the final version of the manuscript. C.E.P. and S.S. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in poster form at the 2023 Pediatric Academic Societies Meeting, Washington, DC, 28 April–1 May 2023.

Handling Editors. The journal editor responsible for overseeing the review of the manuscript was Mark A. Atkinson.