An Observational Study of Maternal Hypoglycemia in Pregnancy

Introduction

Gestational diabetes mellitus (GDM) is a common pregnancy complication, affecting approximately 15% of pregnant individuals globally.¹ GDM poses significant short- and long-term health risks for both the mother and the offspring, emphasizing the importance of early detection and management. In the short term, GDM is associated with maternal and fetal complications such as macrosomia (birth weight >4500 grams), large-for-gestational-age (LGA) infants, increased rates of operative vaginal deliveries and cesarean sections, birth trauma, severe perineal tears, preeclampsia, polyhydramnios, and intrauterine fetal demise. Long-term complications include a heightened risk of metabolic syndrome, cardiovascular disease, and type 2 diabetes mellitus for both mother and child.^2–6

The American College of Obstetricians and Gynecologists (ACOG) outlines that GDM diagnosis involves either a one-step or two-step process, often requiring at least two abnormal glucose values, though a single abnormal value may suffice in specific circumstances. “Abnormal values” are predominantly defined as elevated glucose levels.⁷

Hypoglycemia, defined as plasma glucose levels below 54–70 mg/dL, is a less frequently studied condition. Its implications, particularly in the context of the glucose challenge test (GCT) and oral glucose tolerance test (OGTT), remain a subject of debate. Some studies suggest that individuals with normal OGTT results exhibit comparable maternal and neonatal outcomes to those with hypoglycemia during OGTT. Other studies report a higher prevalence of pathological glucose patterns on self-monitoring, an increased prevalence low-birth-weight (<2500 grams), and a higher likelihood of subsequent GDM diagnosis in individuals with hypoglycemia during OGTT.^8–10

Given that maternal glucose diffuses to the fetus, maternal hypoglycemia could induce fetal hypoglycemia, potentially leading to fetal hypoxia and suboptimal growth. These findings underscore the need for a nuanced understanding of glucose abnormalities during pregnancy and their implications for maternal and neonatal health.¹¹

However, important knowledge gaps remain. Available studies are largely single-center, use heterogeneous hypoglycemia thresholds and testing strategies, and have limited ability to examine whether the timing of hypoglycemia during pregnancy modifies fetal growth risk. In addition, maternal metabolic outcomes beyond pregnancy, including subsequent diabetes risk, have rarely been evaluated in individuals with hypoglycemia detected on screening or diagnostic glucose testing.

The current study aimed to assess the rate of small-for-gestational-age (SGA) and large for gestational age (LGA) neonates among individuals with hyperglycemia, hypoglycemia and normoglycemia. We hypothesized that maternal hypoglycemia will be associated with an increased risk of small-for-gestational-age (SGA) just as hyperglycemia is associated with large for gestational age (LGA) neonates, in addition to pathological maternal metabolic outcomes and neonatal complications, when compared to normoglycemic individuals.

Methods

Study Design

This retrospective analysis of data from electronic medical records of Maccabi Healthcare Services (January 2000–December 2022) is reported according to the Reporting of Studies Conducted Using Observational Routinely Collected Data recommendations.¹²

Setting

The Maccabi Health Service (MHS) is the second-largest health maintenance organization (HMO) in Israel. Founded in 1940, it currently covers approximately 2.4 million insurees. Like all HMOs in the country, the MHS operates according to the guidelines of the national insurance law. The national health insurance law ensures that all residents of the country are entitled to a basket of healthcare services predetermined by a national committee. The costs of all pregnant individuals are mandatorily covered, and this population undergoes scheduled follow-up and testing. Medical treatment follows national guidelines which are written and updated periodically by the Israeli society of obstetrics and gynecology according to international guidelines. Follow-ups and tests are routinely recorded in a standard electronic medical record (EMR). GCT and OGTT are performed according to ACOG recommendations, usually during the second trimester, between 24–28 gestational weeks.

Participants

The MHS database was used to identify those fulfilling the following criteria: Primiparous individuals, aged 18–50 years at the time of their last menstrual period, with a 100-gram oral glucose tolerance test (OGTT) results available, and a live birth. We also included patients with a 50-gram glucose challenge test (GCT) result <70 mg/dl, even if OGTT results were not available. Excluded were individuals with multiple gestation and those with missing data of neonatal outcomes or birthweight. Data were extracted up to a year after delivery.

Variables

The primary study outcome was the rate of large for gestation (LGA) or small for gestation (SGA) neonates. We defined LGA as birthweight >90th centile and SGA as birthweight <10th centile according to local birth charts.¹³ Individuals were classified as hyperglycemic if they had one or more elevated glucose level, as hypoglycemic if they had any glucose levels <70 mg/dl (with or without hyperglycemia) or as normoglycemic.⁷ Analysis was performed according to the trimester during which the pathological glucose level was measured. Since there is no standard definition for GDM, we chose to use the conservative definition of a single pathological value.⁷

The secondary study outcomes were the rates of complications (hypertensive disorder, DM later in life) among individuals with hypoglycemia as compared to those with normoglycemia, and whether these individuals with hypoglycemia were classified as dysglycemic (ie. defined as the presence of any ICD-10 diagnosis related to abnormal glucose metabolism recorded in the patient’s medical chart; the diagnosis was made and documented by the treating physician during the index pregnancy) and required glucose lowering agents. We therefore also collected data on maternal demographics (maternal age at diagnosis, socio-economic status) and co-morbidities (smoking, body mass index [BMI]), the details of the current pregnancy (eg. maternal fetal medicine clinic follow-up, gestational age at diagnosis, glucose test results, gestational age at delivery), pregnancy complications (hypertensive disorders, diagnosis of GDMA1 and GDMA2), medical treatment for GDM and on later presentation of diabetes mellitus (DM according to the diabetes registry).

A list of codes used to identify the variables is provided in Supplementary Table 1.

We also collected data on neonatal outcomes (birthweight, jaundice, hypoglycemia, rates of referral to a pediatrician or the department of emergency medicine during the first year of life, and failure to thrive [FTT]) and data on maternal follow up and outcomes (number of obstetrical medical visits, ultrasound examinations, referrals to the department of emergency medicine, and hospitalizations).

Data Sources/Measurement

All the data were extracted by MHS data analysts from the EMRs of relevant patients identified in the MHS database. All MHS tests are sent to standardized ISO-accredited central laboratories and results are sent electronically directly to the treating clinicians as well as being automatically documented in the EMRs. Measurements were unrelated to the study and were performed by independent clinicians and laboratory workers.

Bias

Selection bias is less likely given that MHS is the second-largest HMO and covers almost a third of the population in the country. Self-selection for testing is always a possibility despite full coverage of costs. Measurement bias is not likely given the rigorous standards upheld by the central MHS laboratories. Some doctors may have failed to document the presence of diabetes in the list of diagnoses. We overcame this issue by classifying patients according to their blood test results rather than through these lists.

Study Size

Study power was not calculated since this study covered the full database population rather than a sample.

Quantitative Variables

We studied age both as a continuous variable and as age above and below 35 years, due to the American Diabetes Association recommendation to screen for DM at this age in populations at risk.¹⁴ Blood pressure was classified according to the American Heart Association classification of hypertension.¹⁵ Neonatal weight categories and LGA and SGA were defined as commonly accepted.

Statistical methods

Missingness is reported. Categorical data were reported as the number (%) and compared using the Chi-square test. Continuous variables were reported as mean ± SD and median [Interquartile Range (Q1-Q3]]. Between-group comparisons were performed using analysis of variance (ANOVA) for normally distributed data and the Kruskal–Wallis test for non-parametric data. Logistic regression analysis was used to examine neonatal and maternal outcomes. All models were adjusted for maternal age at diagnosis, dysglycemia and GDMA2 unless GDMA2 was studied as an outcome. Models studying GDMA2 as an outcome were adjusted only for maternal age at diagnosis and dysglycemia. Models studying dysglycemia as an outcome were adjusted for maternal age at diagnosis and GDMA2. All data were analyzed using SPSS version 29 (IBM Corp. Released 2023. IBM SPSS Statistics for Windows, Version 29.0.2.0 Armonk, NY: IBM Corp). P-values less than 0.05 were considered significant.

Data Access and Cleaning Methods

MHS provides data analysts and statistical support. Aggregate data are provided to the researchers following data cleaning. Data cleaning included the following: OGTT with the value of “0” were considered missing, illogical neonatal weights (eg. 6666, 38). All analyses are performed by the support team as per request.

Linkage

Maternal and neonatal files were linked (person level) across a single data base using IDs with a proprietary code unique to the MHS.

Results

Participants

Overall, 169,346 potentially eligible primiparous individuals aged 18–50 years at the time of their last menstrual period with a pregnancy resulting in a live birth were identified in the MHS database. After excluding individuals with incomplete 100-gram oral glucose tolerance test (OGTT) and adding individuals who had hypoglycemic value (<70 mg/dl) in their 50-gram glucose challenge test (GCT), there were 47,086 eligible individuals. After further exclusion for multiple gestation and missing data regarding neonatal outcome or neonatal weight, the final study cohort included 39,306 individuals (Figure 1).

Figure 1 Inclusion exclusion flow chart.

Missingness

Variables with the greatest amount of missing data were BMI in early pregnancy (84%), BMI at delivery (90%), weight in early pregnancy (60%), weight at delivery (53%) and fasting glucose in early pregnancy (50%).

Descriptive Data

The average maternal age at the time of diagnosis ranged between 28–31 years regardless of trimester of diagnosis. Diagnostic tests were performed during first, second and third trimester at an average of 9, 25 and 32 weeks of gestation, respectively. Most individuals were diagnosed during the second trimester of pregnancy. Additional maternal and neonatal characteristics are presented in Table 1.

Table 1 Population Characteristics

Outcome Data

Overall, 8543 patients had hyperglycemia, 16,062 had hypoglycemia and 14,701 were normoglycemic.

Main Results

When comparing patients with hyperglycemia to normoglycemia in the second trimester, the likelihood of SGA <10 centile was 1.2 (95% CI 1.1–1.4), and the likelihood of SGA <1 centile was 3.5 (95% CI 1.6–7.8). This difference was not apparent in the third trimester.

When comparing patients with hypoglycemia to normoglycemia, the likelihood of SGA <10 centile was 1.2 (95% CI 1.1–1.4) in both the second and the third trimester, and the likelihood of SGA <1 centile was 2.5 (1.1–5.3) in the third trimester only. (Table 2). Neonatal follow-up during the first year of life after maternal hyperglycemia, hypoglycemia and normoglycemia is presented in Supplementary Table 2 per trimester of diagnosis.

Table 2 Neonatal Outcomes

Compared to individuals with normoglycemia the likelihood of hypertensive disorder was increased in both hyper- and hypoglycemia. However, the number of patients in each group was very small. Compared to individuals with normoglycemia, individuals with hypoglycemia were less likely than individuals with hyperglycemia to be diagnosed with DM later in life. Compared to individuals with normoglycemia, the likelihood of diagnosis of dysglycemia during pregnancy (had ICD code for gestational diabetes in their medical chart) was also lower in individuals with hypoglycemia versus hyperglycemia. Although many hypoglycemic patients received treatment with glucose lowering agents (GDMA2), the adjusted likelihood of such treatment was lower in individuals with hypoglycemia versus hyperglycemia when compared to normoglycemia. (Table 3)

Table 3 Maternal Outcomes

Utilization of obstetric services among patients with hyperglycemia, hypoglycemia and normoglycemia per trimester of diagnosis is presented in Supplementary Table 3.

Discussion

Principal Findings

In this observational study of nearly 40,000 pregnant individuals who underwent either a GCT or an OGTT, the likelihood of SGA neonates was higher among those with hypoglycemia than those with normoglycemia. This association was most pronounced in the second and third trimesters of pregnancy. However, neonatal follow-up revealed no clinically significant differences in the studied variables within the first year of life.

Regarding maternal outcomes, the prevalence of hypertensive disorders was slightly higher in individuals with either hyperglycemia or hypoglycemia versus normoglycemia, although the difference appeared negligible. As expected, individuals in hyperglycemia group had a higher prevalence of GDMA2, more individuals were diagnosed with dysglycemia during pregnancy (presence of ICD-10 diagnosis related to abnormal glucose metabolism recorded in the patient’s medical chart by the treating physician during the index pregnancy) and more individuals were diagnosed with DM later in life, across all three trimesters. Interestingly, individuals with hypoglycemia also had a higher prevalence of GDMA2 and dysglycemia, albeit a similar prevalence of DM later in life, compared to individuals with normoglycemia. Due to the retrospective design of the study, the exact mechanism underlying this relationship remains unclear, but it is possible that patients with hypoglycemia underwent self-glucose monitoring more frequently, ultimately revealing previously undiagnosed diabetes requiring medical intervention. This finding supports the hypothesis that hypoglycemia following a glucose challenge may be an unrecognized marker of diabetes. Nevertheless, individuals with hypoglycemia versus hyperglycemia were less likely to be diagnosed with DM later in life. Additionally, while individuals with hyperglycemia had more frequent obstetric visits, those with hypoglycemia did not differ significantly in their visit frequency compared to normoglycemic individuals.

Results in Context of What is Known

Our study is unique in that it included a somewhat larger cohort than the largest study to date.¹⁶ Of previous studies evaluating hypoglycemia and pregnancy outcomes, two were conducted at the UK,^10,17 one in China,¹⁸ one in the US¹⁶ and one in Israel.⁸ Four studies included data from a single medical center^8,10,16,17 but only one included community-level data such as ours.¹⁸ Although two reports described maternal outcomes,^16,17 these included pregnancy-related outcomes only, like hypertensive disorders, mode of delivery, postpartum hemorrhage and diagnosis of GDM. In the current study, we also examined the rate of diagnosis of DM later in life.

Our study included only individuals with a live birth as did prior studies.^10,16–18 Our findings that hypoglycemia is associated with SGA were similar to several previous studies.^10,16,18 Raviv et al found that the rate of SGA was similar once adjusted to maternal age, BMI, nulliparity, and any hypertensive disorder. However, they also stated that “it would be hard to ignore the differences in birthweight trends”.⁸

We found no clinically significant difference between hyperglycemic, hypoglycemic and normoglycemic individuals in either age or the rate of hypertensive disorders and therefore did not adjust for these variables, and the paucity of data regarding BMI in our database did not enable adjustment for this variable. Furthermore, our study population also differed from that of Raviv et al in maternal age (lower percent of individuals >35 years) and rate of nulliparity (100% vs. 30%).

Contrary to most prior studies,^10,16,17 we limited the upper age of the included population to 50 years. Leng et al¹⁸ included only individuals aged less than 39 years Our study included only individuals in their first pregnancy whereas the rest included both nulliparous and multiparous individuals.^10,16–18 Across all studies, including ours, the criteria for inclusion in terms of glucose testing that were used for diagnosis and the cutoffs used for classification of abnormality were inconsistent. For example, Ding et al used a cutoff of 82 mg/dl to define hypoglycemia during 50 g GCT,¹⁶ Nayak et al¹⁰ used 88.5 mg/dl, while we defined hypoglycemia less than 70 mg/dl on either 50 g GCT or 100OGTT.

Clinical Implications

In an ideal world, glucose tolerance testing would be conducted in the first trimester of pregnancy. Already at this early stage, more than half of individuals had abnormal glucose values. Individuals with a normal test during first trimester should repeat testing between 24–28 weeks where the highest insulin resistance, while individuals with abnormal glucose, high or low, should be requested to keep a glucose level diary for 1–2 weeks. Those with entirely normal values or sporadic hyperglycemia or hypoglycemia could probably continue intermittent self-glucose monitoring with appropriate lifestyle modification and regular maternal follow up. Those with abnormal values (> 15–20%) despite lifestyle modification, should be referred to maternal fetal medicine specialist clinics to consider further medical treatment. Glucose values should be addressed as a continuum, with pathology being recognized by consistently irregular findings on both sides of the normal spectrum.

Research Implications

Future studies should consider testing the feasibility of such an approach, should include highly granular, prospective data to enable controlling for potential confounders in a sufficiently large sample and should also include more outcome data on patient-centered outcomes.

Strengths and Limitations

Strengths of this study include its large, population-based sample size and the inclusion of long-term maternal outcomes—an aspect rarely addressed in prior work. The exclusive inclusion of primiparous individuals helps control for confounding due to obstetric history. Moreover, the use of real-world data from a national health maintenance organization (HMO) enhances generalizability.

The current study was retrospective, and therefore, despite the large sample size, determines association rather than causation. Although the MHS database covers almost one-third of the population in the country, this sample may still not represent the entire population, nor may it represent pregnant population elsewhere. The need to ensure generalizability requires further research. Our criteria for study inclusion were somewhat different from prior studies. However, the criteria for diagnosis of hypoglycemia during pregnancy are inconsistent across multiple papers. Due to limitations in file linkage, our data does not include in-hospital data such as complications during labor, mode of delivery, Apgar score, umbilical cord pH and neonatal intensive care unit admission.

Conclusion

Maternal hypoglycemia during pregnancy is associated with an increased risk of SGA neonates and may serve as an early marker of underlying glucose dysregulation. Given these findings, repeated glucose monitoring and closer follow-up for individuals with hypoglycemia may be warranted. Several considerations should guide interpretation of these findings. First, given the retrospective observational design, the associations observed cannot establish causality. Prospective validation is also required. Second, the definition of hypoglycemia in pregnancy may need further refinement.

Future studies should include prospective, trimester-specific assessment of glucose dynamics with standardized thresholds, incorporate detailed maternal metabolic phenotyping and nutritional data, and evaluate longer-term maternal and neonatal metabolic implications and outcomes. Such work is needed to inform evidence-based clinical management strategies.