Hemoglobin A_1c Trajectories During Pregnancy and Adverse Outcomes in Women With Type 2 Diabetes: A Danish National Population-Based Cohort Study Free

Compared with pregnant women with normal glucose tolerance, women with type 2 diabetes have an increased risk of pregnancy complications, including perinatal deaths, congenital malformations (CM), preterm deliveries, caesarean sections (C-section), and extreme birth weights, emphasizing the severity of the condition (1–5). Appropriate glycemic control is fundamental to decreasing the risks of severe adverse outcomes (2). Therefore, pregnancy poses a critical therapeutic challenge, as insulin resistance increases considerably in maternal tissues, resulting in severe insulin resistance in some women with diabetes (6).

During pregnancy, self-monitoring of capillary blood glucose is the primary measure of glycemic control, guiding insulin treatment (7). However, glycated hemoglobin (HbA_1c) reflects long-term glycemic control and is used as a prognostic biomarker, influencing the clinical decisions (7). Various thresholds for optimal HbA_1c before and during pregnancy have been proposed, but few studies have accounted for potential differences in appropriate HbA_1c between type 1 and type 2 diabetes (7). Furthermore, the ability of women with type 2 diabetes to improve glycemic control after conception has not been explored. This knowledge is needed, as previous studies have shown a low attendance for prepregnancy care (2,5), reflected in a high prevalence of women with type 2 diabetes entering pregnancy with marked hyperglycemia (2).

For the women and their caregivers, it is of the utmost importance to have insight into the potential for improving glycemic control after conception. This may be explored by latent class linear mixed-model (LCLMM) analysis, as this method reveals unspecified groups within a cohort according to changes in exposure across time and the associated risks. LCLMM has been used to identify HbA_1c trajectories outside pregnancy in previous studies (8,9), but it has never been done during pregnancy.

This study aimed to identify and characterize groups of pregnant women with type 2 diabetes with distinct HbA_1c trajectories across gestation and to examine the association between such trajectories and adverse obstetric and perinatal outcomes. We hypothesized that the HbA_1c trajectories would reveal an improvement potential determined by the degree of hyperglycemia at conception, resulting in a difference in the odds of severe adverse outcomes.

This retrospective Danish nationwide population-based cohort study included all identifiable singleton pregnancies in women with preexisting type 2 diabetes, delivering at one of the four University Hospitals between 1 January 2004 and 14 November 2019. These four hospitals were charged with antenatal and perinatal care of all women with type 2 diabetes in Denmark during the study period. Participants were excluded if they did not have a confirmed diagnosis of type 2 diabetes before conception, had other types of diabetes, had a twin pregnancy, were missing a patient record, did not deliver a liveborn infant after 24 weeks of gestation, or had less than three HbA_1c measurements during pregnancy. Women could contribute with consecutive pregnancies during the study period.

The identification of pregnancies in women with type 2 diabetes was based on the ICD-10 coding system (DO241). Electronic and paper-based medical records were reviewed to validate the study population and collect data.

Data were collected on the following variables for population characterization: maternal age, parity, ethnic affiliation, smoking, chronic comorbidities, duration of diabetes, treatment of diabetes, prepregnancy HbA_1c (last measurement registered within 12 weeks before conception), prepregnancy BMI, and weight change during pregnancy in kilograms. Insulin dose was reported as maximum total IU/day according to trimester and maximum total IU/kg/day during pregnancy. Oral antidiabetes treatment was not used during pregnancy. Prepregnancy weight, smoking, and ethnicity were self-reported by the women.

According to Danish guidelines, HbA_1c targets were <7.0% (53 mmol/mol) before pregnancy, <6.5% (48 mmol/mol) in early pregnancy, and <5.6% (38 mmol/mol) in late pregnancy (10). Weight gain during pregnancy was categorized according to Danish guidelines as inadequate, appropriate, or excessive: women with a BMI <25.0 kg/m² were recommended to gain 10–15 kg; women with a BMI 25.0–29.9 kg/m², 5–8 kg; and women with a BMI ≥30 kg/m², 0–5 kg (10).

The primary exposure of interest was HbA_1c trajectories across gestation. All measurements of HbA_1c registered in the medical records from conception until delivery were included in the analyses.

The primary outcome was standardized infant birth weight adjusted for fetal sex and gestational age at birth, reported as z scores using the method by Marsal et al. (11), and categorized as the two birth weight extremes: small for gestational age (SGA) (<10th percentile) and large for gestational age (LGA) (>90th percentile).

Secondary outcomes included adverse events during pregnancy and delivery: maternal complications (hypoglycemia resulting in hospital admittance, pregnancy-induced hypertension, preeclampsia), gestational age at birth, preterm birth (<32 and <37 weeks of gestation), C-section, CM, and neonatal complications (hypoglycemia and jaundice within 24 h, mortality within 7 days). The clinicians registered pregnancy-induced hypertension and preeclampsia according to Danish guidelines at the time. CM was classified according to the ICD-10 and included minor and major CM diagnosed at fetal ultrasound during pregnancy or during the postpartum hospital stay.

Longitudinal trajectories of HbA_1c were modeled using LCLMM analysis as a function of gestational age specified with restricted natural cubic splines. Knots were placed at the limits and quartiles of gestational age. Individual-specific random intercepts were included in the model to account for the longitudinal nature of the data. The number of groups was not prespecified but determined by evaluating two to five group model solutions according to posterior group membership probability (probability of the individuals in the cohort belonging to the assigned group), Akaike information criterion, Bayesian information criterion, relative entropy (degree of group separation), and group sizes. Furthermore, the final model was evaluated according to the clinical relevance of the trajectories. Model-estimated mean HbA_1c values with SEs were reported at five evenly distributed time points across gestation, including differences between groups.

Variables are reported as mean and SD for continuous variables with normal distribution (examined by quantile-quantile plots), median and interquartile range (IQR) for continuous variables with skewed distribution, and count and percentage for categorical variables. Associations between groups and severe adverse outcomes were analyzed using logistic regression models (LRMs) with robust clustered SEs to account for women contributing multiple deliveries and being nested within different hospitals. LRMs were adjusted for prepregnancy BMI and weight gain during pregnancy (model 1), and BMI, weight gain, maternal age, duration of diabetes, ethnicity, parity, smoking, and chronic hypertension (model 2). BMI and weight gain were examined separately, as they were hypothesized to have a strong confounding effect. Results are reported as the crude and adjusted odds ratios with 95% CIs and SEs. A complete case analysis was conducted. Data were analyzed using StataSE 17.0 software (StataCorp Ltd, College Station, TX), and estimations of LCLMM were performed using the lcmm 2.0.2 package in R 4.3.0 (R Foundation for Statistical Computing, Vienna, Austria).

The study was approved by the Danish Society for Patient Safety (jr. 31-1521-33), Frederiksberg, Copenhagen, Denmark and the Regional Council in the Central Denmark Region (jr. 1-45-70-52-20), Viborg, Denmark.

Data are not publicly available according to the General Data Protection Regulation in Denmark. However, they are available from the corresponding author upon reasonable request and with permission of the Regional Council in the Central Denmark Region.

A total of 1,129 pregnancies met the inclusion criteria (Supplementary Fig. 2). Using LCLMM analysis, a three-group model with distinct trajectories of HbA_1c across gestation was chosen (Fig. 1, Supplementary Tables 4–6, and Supplementary Fig. 3). Group names were assigned according to the glycemic control in early pregnancy, where the between-group variability was most pronounced: 671 (59.4%) were included in the good glycemic control group, 357 (31.6%) in the moderate glycemic control group, and 101 (8.9%) in the poor glycemic control group. The mean probability of belonging to the assigned group was 91.7%, 89.3%, and 94.3% for the three groups, respectively (Supplementary Table 4). The relative entropy for the model was 0.82.

According to the definition, the group with good glycemic control had the lowest estimated mean HbA_1c at 6.4% (46.2 mmol/mol) at 30 days of gestation compared with 8.1% (65.4 mmol/mol) in the group with moderate glycemic control, and 11.1% (98.3 mmol/mol) in the group with poor glycemic control (Supplementary Table 4). All groups initially improved their HbA_1c, with the poor glycemic control group having the steepest decline (Fig. 1). No difference in HbA_1c was found between the good and moderate glycemic control groups from 168 days of gestation (24 weeks) and the good and poor glycemic control groups from 199 days of gestation (28 weeks).

The model did not identify a subgroup with increasing levels of HbA_1c or a steady high HbA_1c during pregnancy. In Danish guidelines, treatment failure is defined as a change in HbA_1c <2.6% (5 mmol/mol). Using this cutoff to define a steady HbA_1c, 56.7% had a steady HbA_1c in the good glycemic control group compared with 6.7% in the moderate glycemic control group and none in the poor glycemic control group. Only women from the good glycemic control group had an increase in HbA_1c during pregnancy (16.2%), and in this group, median (IQR) HbA_1c was 6% (5.6– 6.4) (42 [38–46] mmol/mol) at the first measurement compared with 6.8% (6.5–7.5) (51 [47–58] mmol/mol) at the last.

Women in the moderate glycemic control group with a steady HbA_1c (6.7%) had a median (IQR) HbA_1c at 6.3% (6.1–6.7) (45 [43–49.5] mmol/mol) at the first measurement compared with 6.1% (5.8–6.5) (43 [39.5–47] mmol/mol) at the last. This small group did not have a strong group membership. In the poor glycemic control group, 16.8% had an HbA_1c >6.7% (50 mmol/mol) at the last measurement in the 3rd trimester, with median HbA_1c at 10.8% (9.7–11.3) (94 [83–100] mmol/mol) at the first measurement compared with 7.5% (7.0–8.2) (58 [53–66] mmol/mol) at the last.

The study population was characterized by a high maternal age, a diverse ethnic affiliation, and a high prevalence of smoking and multimorbidity. Most women in all three groups were obese and had an excessive weight gain during pregnancy (Table 1).

Expectedly, more women had a prepregnancy HbA_1c <7.0% (53 mmol/mol) in the good glycemic control group compared with the moderate and poor glycemic control groups (75.3% vs. 31.8% vs. 18.2%, respectively). However, prepregnancy HbA_1c was only available for 38.0%, 37.0%, and 21.8% in the three groups, respectively. Very few attained the recommended HbA_1c target of <5.6% (38 mmol/mol) in late pregnancy, with no differences between the groups (21.0% vs. 22.5% vs. 21.7%, respectively) (Table 1).

Most women in all groups needed insulin during pregnancy. The good glycemic control group received a median (IQR) daily insulin dose during the 3rd trimester at 82 (46–140) IU/day compared with 112 (67–173) in the moderate control group and 129 (84–198) in the poor glycemic control group (Table 1).

The mean (SD) birth weight z score was 0.8 (1.7) in the good glycemic control group, 0.7 (1.6) in the moderate glycemic control group, and 0.4 (1.7) in the poor glycemic control group (Table 2). The prevalence of LGA infants was 35.1%, 31.3%, and 30.0% in the good, moderate, and poor glycemic control groups, and the prevalence of SGA infants was 9.4%, 8.5%, and 19.0%, respectively (Table 2). The LRM analysis used the good glycemic control group as a reference. The poor glycemic control group had 0.57 lower adjusted odds of having an LGA infant (95% CI 0.40–0.83) and 2.49 higher adjusted odds of having an SGA infant (95% CI 2.00–3.10). No difference was found between the moderate and good glycemic control groups in the odds of having an LGA infant after adjusting for BMI and weight gain, and no difference was found in the odds of having an SGA infant (Table 3).

In addition, the poor glycemic control group had 4.60 higher adjusted odds of having an infant with a CM (95% CI 3.39–6.26); whereas, no evidence of a difference was found between the moderate and good glycemic control groups (Table 3). The odds of having a C-section and a composite outcome were higher in the poor glycemic control group in the unadjusted models, but no differences were observed after adjusting for BMI and weight gain. In the total study population, 73.0% of pregnancies had at least one adverse event defined as the composite outcome. No evidence of a difference in the odds of preeclampsia or preterm birth was found between the groups (Table 3).

In this nationwide cohort study of pregnancies complicated by maternal type 2 diabetes, three distinct groups of HbA_1c trajectories across gestation were identified. The groups were characterized by varying glycemic control in early pregnancy. Still, from midpregnancy onward, all groups had an estimated HbA_1c <6.5% (48 mmol/mol), with no differences between groups in the 3rd trimester. Women belonging to the poor glycemic control group had increased odds of having an infant with SGA birth weight and CM compared with women belonging to the good glycemic control group, but lowered odds of having an infant with LGA birth weight. Still, the prevalence of LGA infants was high in all three groups, and an explanation may be that few women in the three groups attained the recommended HbA_1c target in late pregnancy.

As LCLMM analysis reveals unspecified groups based on the data input, it may provide a more nuanced understanding of the ability of pregnant women with type 2 diabetes to improve HbA_1c levels during pregnancy and the concomitant association with adverse outcomes during pregnancy.

LGA infants are prevalent in pregnant women with type 2 diabetes (2), observed in one in three infants in the current study. Notably, the point estimates showed the highest prevalence of LGA infants in the good glycemic control group, whereas the poor glycemic control group had the highest prevalence of SGA infants. These extreme birth weights from both ends of the scale showed strong associations with the HbA_1c trajectory group.

A previous study from one of our hospitals observed a positive association between high HbA_1c in early pregnancy and low adjusted birth weight in women with type 1 diabetes (12). Unfortunately, risk factors for SGA birth weight in women with type 2 diabetes have been scarcely examined, with no consensus on the association with HbA_1c (13–15). The risk of SGA infants is often not considered in pregnant women with diabetes. Still, infants of women with type 2 diabetes have been shown to have a higher risk of SGA birth weight compared with infants of women with type 1 diabetes (2,16). Whether the risk is higher than in women without diabetes remains uncertain (14,16). In the current study, 19.0% of infants in the poor glycemic control group were born SGA, which is more than expected according to the reference population (11). A focus on both SGA and LGA infants is important, as both are associated with severe short- and long-term childhood morbidities, including obesity, diabetes, and cardiovascular disease (17).

The pathophysiological mechanisms linking hyperglycemia to both birth weight extremes may be that hyperglycemia during early pregnancy affects placentation (18), limiting fetal growth in late pregnancy (19). In contrast, hyperglycemia in late pregnancy contributes a high nutritional supply to the fetus, increasing the risk of fetal adiposity (20). This indicates that changes in glycemic control at different critical periods during pregnancy might affect fetal growth and development differently. Accordingly, Glinianaia et al. (15) found an accelerated effect of increasing 3rd trimester HbA_1c on the risk of high birth weight when the HbA_1c at conception was low. Furthermore, a study by Hauffe et al. (21) observed that achieving 1st trimester HbA_1c <6.5% (48 mmol/mol) did not lower the risk of LGA infants if the 3rd trimester HbA_1c was >6.0% (42 mmol/mol). This could explain the higher prevalence of LGA infants in the good glycemic control group compared with the other groups.

Results from previous studies on an association between HbA_1c and LGA infants have been conflicting. As mentioned, Glinianaia et al. (15) and Hauffe et al. (21) found an association between higher 3rd trimester HbA_1c and increased risk of LGA infants in women with preexisting diabetes, and the same was reported in women with type 2 diabetes by Murphy et al. (2). By contrast, two studies, by Abell et al. (14) and Ladfors et al. (22) found no association between HbA_1c and LGA infants in women with type 2 diabetes. However, the studies included small, well-regulated populations with a low prevalence of LGA infants (14,22), and Abell et al. (14) only examined a single mean HbA_1c during pregnancy.

In the current study, the odds of having an infant with a CM in the poor glycemic control group were 1 in 7, compared with 1 in 19 in the moderate glycemic control group and 1 in 27 in the good glycemic control group. Data from an appropriate control population without diabetes was not available. However, a register-based study including a subset of this cohort found 90% increased odds of CM in offspring of women with type 2 diabetes compared with unexposed offspring (adjusted odds ratio 1.9, 95% CI 2.8–1.3) (23). Including women with type 1 diabetes, the risk of CM increased exponentially with increasing HbA_1c levels in early pregnancy, and the risk was still higher in offspring of women with diabetes than in unexposed offspring at HbA_1c levels <6.5% (48 mmol/mol) (23). Unexpectedly, the association between HbA_1c in early pregnancy and risk of CM was not as strong using data from women with type 2 diabetes only in the register-based study. However, this may be explained by the size of the study population and the few cases of CM (23). Abell et al. (14) found no association between HbA_1c and risk of CM in women with type 2 diabetes. However, as mentioned, they included a small, well-regulated population, and there were few cases of CM (14).

In comparison, a cohort study of 8,684 women with type 2 diabetes by Murphy et al. (2) found the level of 1st trimester HbA_1c to be the strongest independent risk factor for developing CM. This is consistent with the results in the current study, as the groups primarily differed in glycemic control in early pregnancy. It also aligns with the organogenesis occurring in gestational weeks 3 to 8, which may be impacted by hyperglycemia through induction of oxidative stress, resulting in embryonic damage (24). However, the results of the current study also suggest that women with good and moderate glycemic control may have similar odds of having an infant with CM.

Other previous studies examining a combined group of women with preexisting diabetes also found an association between HbA_1c in early pregnancy and risk of CM (3,25–29). Notably, a register-based study by Bell et al. (3) found a linear association between the risk of CM and the level of periconceptional HbA_1c, with no specific threshold. However, a study by Davidson et al. (27) found that women were able to lower their risk of CM if the HbA_1c levels were improved during early pregnancy until gestational week 21. Interestingly, after stratifying for prepregnancy HbA_1c, the positive effect of reducing HbA_1c in early pregnancy was only present in women with a prepregnancy HbA_1c ≥6.4% (46 mmol/mol) (27). This may explain why women with good and moderate glycemic control had similar odds of having an infant with CM in the current study.

In the current study, preeclampsia and preterm birth were not associated with the HbA_1c trajectory group. In women with type 1 diabetes, HbA_1c is a known risk factor for preeclampsia and preterm birth (30,31), but it is still uncertain in women with type 2 diabetes (2,14,21,32). C-section was associated with the HbA_1c trajectory group in the unadjusted model but not after adjusting for BMI and weight gain. Previous studies found no association between HbA_1c during pregnancy and C-section (14,33,34). Instead, C-section was associated with shorter maternal height, which may be conveyed through higher BMI (34).

As pregnancy progresses, the interpretation of HbA_1c becomes more complicated. HbA_1c is formed continuously by modifying hemoglobin molecules according to the blood glucose concentration. During pregnancy, the relative life span of the erythrocytes is gradually reduced due to increased erythropoiesis and higher cellular turnover, resulting in a reduction in HbA_1c (35). In addition, the concentration of hemoglobin is decreased due to hemodilution (35). Studies have shown, that the upper limit of HbA_1c is lower in pregnant women without diabetes compared with nonpregnant women (36–38). HbA_1c may also be lower in the 2nd compared with the 1st trimester (36,37) and increase in the 3rd trimester (36,38). This may explain why additional improvement in late pregnancy is not easily achieved. Notably, the inter- and intraindividual changes in HbA_1c may not only be attributed to differences in glycemic control.

The inconsistent results in the literature regarding the significance of HbA_1c during pregnancy in women with type 2 diabetes may question its value as a prognostic biomarker. However, the results from the current study suggest that HbA_1c measurements in early pregnancy are relevant to establishing the risk of having an infant with extreme birth weight and CM, which may be reflective of placental dysfunction (18,19). Furthermore, previous studies have focused a lot on prepregnancy HbA_1c. However, as reported in the current study, few women have an available HbA_1c value measured within 3 months before conception, which may reflect poor attendance for prepregnancy care (2,5). Therefore, we need to be able to stratify the women according to HbA_1c measurements during pregnancy and understand the associated risks of adverse outcomes. Using LCLMM analysis, we have described the expected HbA_1c trajectory during pregnancy in women with type 2 diabetes in a real-world clinical setting. In addition, we have examined the concomitant associations with adverse outcomes, which may help guide clinicians.

Some strengths and weaknesses related to the current study should be considered. The retrospective design meant we could not obtain data on postprandial glucose values, which could have contributed to an alternative measure of glycemia. Data were not included on pregnancies terminated before gestational week 24 or stillborn offspring delivered after gestational week 20, which may have resulted in an underestimation of the prevalence of CM. Furthermore, data were restricted to CM diagnosed antenatally or during the postpartum hospital stay, which may have resulted in an underestimation of the association with HbA_1c. There were no neonatal deaths, preventing us from commenting on this important outcome. Excluding pregnancies with less than three HbA_1c measurements could have introduced selection bias. Finally, individual pregnancies may have a probability of belonging to any group identified by LCLMM analysis, which was not appraised after group assignment. However, in the current study, the probability of belonging to the assigned group was high.

In conclusion, the current study demonstrated that women with type 2 diabetes with severe dysregulated diabetes in early pregnancy could improve HbA_1c values toward target levels within the first half of pregnancy, thereby attaining the same level of glycemic control as women with well-regulated diabetes. Still, the odds of infants with SGA birth weight and CM were higher compared with women with less pronounced HbA_1c trajectories. Concerningly, the prevalence of LGA infants was highest among women in the good glycemic control group. Future studies should explore whether HbA_1c measured in late pregnancy is an appropriate prognostic biomarker for adverse perinatal outcomes, especially in women with seemingly well-regulated diabetes.

Funding. This work was funded by a Novo Nordisk Foundation Steno Collaborative Grant 2018 (grant no. 0052373). Steno Diabetes Center Aarhus is partly funded by a donation from the Novo Nordisk Foundation (no. NNF17SA0031230). A.H. is supported by a Novo Nordisk Foundation Data Science Emerging Investigator grant (no. NNF22OC0076725).

The funders were not involved in the work reported in this paper, including study design, data collection, data analyses, preparation of manuscripts, or publication decisions.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.S.K. wrote the first draft of the manuscript. A.S.K. collected the data. A.S.K., S.K., M.L.-M., and A.H. interpreted the results. A.S.K. and M.L.-M. performed the data analysis. S.K. and A.H. verified the analytical methods. A.E.R. contributed with data. All authors were involved in the conception and design of the study. All authors edited, reviewed, and approved the final version. A.S.K. is the guarantor of this work and, as such, has full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Cheryl A.M. Anderson and Cuilin Zhang.