Novel Biochemical Markers of Glycemia to Predict Pregnancy Outcomes in Women With Type 1 Diabetes Free

Type 1 diabetes in pregnancy is associated with obstetric and neonatal complications, which are attributed to maternal hyperglycemia (1). Affected women are more likely to develop preeclampsia and to experience instrumental or operative deliveries (1). Their newborn babies may be affected by preterm delivery, large for gestational age (LGA) (birth weight >90th centile), and neonatal hypoglycemia, which contribute to high rates of admission to the neonatal intensive care unit (NICU) (1).

Neonatal complications of type 1 diabetes pregnancy can be prevented or ameliorated by improved maternal glucose levels (2,3). However, the objective assessment of maternal glycemia throughout pregnancy is challenging. Gestational changes in red cell turnover and serum protein concentrations raise concerns about the validity of HbA_1c as a glycemic marker (4,5). HbA_1c measurements typically reflect glycemia over the preceding 2–3 months, which is also less suitable for intensive monitoring of 1–2 weekly glucose patterns during pregnancy (5).

Novel approaches to the assessment of glycemia in pregnancy include the use of continuous glucose monitoring (CGM) metrics and alternative laboratory markers, including glycated CD59 (gCD59), 1,5-anhydroglucitol (1,5-AG), fructosamine, and glycated albumin (6–8). During the Continuous Glucose Monitoring in Women With Type 1 Diabetes in Pregnancy Trial (CONCEPTT), the use of CGM led to improved maternal glycemia and neonatal outcomes, with a substantial reduction in admissions to the NICU (9). This has resulted in widespread adoption of CGM in pregnant women with type 1 diabetes. Many women now have CGM data available throughout pregnancy, but it is unclear which metrics, and at which time points, are most useful for pregnancy outcome prediction. Recent work has identified that lower time in range (TIR), higher mean glucose, and glucose SD are associated with LGA (10). Functional analysis of CGM data from CONCEPTT participants has identified that women whose offspring develop LGA have a higher mean glucose for 14–16 h per day throughout pregnancy (11). The aim of the current study was to assess the predictive performance of HbA_1c, CGM metrics, and alternative biochemical markers to identify women with type 1 diabetes at risk for suboptimal pregnancy outcomes.

CONCEPTT included women with type 1 diabetes recruited during pregnancy or while planning pregnancy and is described elsewhere (9). The trial received ethical approval from the Health Research Authority, East of England Research Ethics Committee, Cambridge, U.K. (12/EE/0310), for all U.K. sites and at each individual center for all other sites. All participants gave written informed consent. The current study is a prespecified secondary analysis approved by the CONCEPTT Steering Committee before trial completion.

Women were randomized to real-time CGM (Guardian REAL-Time or MiniMed Minilink system; Medtronic, Northridge, CA) in addition to capillary glucose monitoring or capillary glucose monitoring alone for diabetes management. Women in the capillary glucose monitoring group also had short periods of masked CGM (iPro2 Professional CGM; Medtronic) to allow comparison of CGM metrics between groups. This study includes data from pregnant and prepregnant recruits who became pregnant and gave birth to a liveborn infant. Prespecified obstetric and neonatal outcomes are preeclampsia (systolic blood pressure of ≥140 mmHg and/or a diastolic blood pressure ≥90 mmHg on two or more occasions a minimum of 6 h apart and proteinuria of ≥1+ on dipstick or ≥300 mg per 24 h), preterm delivery (<37 weeks), LGA (>90th centile) on the basis of customized centiles, neonatal hypoglycemia requiring intravenous dextrose, and NICU admission requiring a duration of at least 24 h.

The HbA_1c samples obtained at ∼10–12, 24–25, and 34–35 weeks’ gestation were shipped at the end of pregnancy and were unavailable to participants and health care teams during the trial. HbA_1c measurements were done using the turbidimetric inhibition immunoassay for hemolyzed whole blood on the COBAS INTEGRA 700 platform (Roche, Basel, Switzerland) at a central laboratory (DynaCare, Brampton, Ontario, Canada). Women were asked to give a voluntary additional serum sample for metabolic studies at the same three time points as trial HbA_1c sampling. The sample was processed quickly and aliquoted for storage at −80°C.

Analyses for 1,5-AG, fructosamine, and glycated albumin were performed in batches in the Core Biochemical Assay Laboratory on the Cambridge Biomedical Campus. 1,5-AG was measured using commercially available reagents (GlycoMark, New York, NY) on an RX daytona+ analyzer (Randox Laboratories, Crumlin, U.K.). Fructosamine was measured using a Randox kit on the RX daytona+ analyzer. Glycated albumin was measured using commercially available reagents from Asahi Kasei Pharma (Tokyo, Japan) on an RX daytona+ analyzer. Analyses for gCD59 were performed in the Laboratory for Translational Research, Harvard Medical School, as described previously (12).

The CGM metrics were derived from periods of 6 days at ∼10–12, 24–25, and 34–35 weeks’ gestation. Definitions were consistent with international reporting recommendations (13) and are defined in Supplementary Table 1.

Continuous data were described as mean (SD) and categorical data as n (%), as appropriate. Pearson correlation coefficients were used to assess relationships between CGM metrics and laboratory markers of glycemia. Unadjusted standardized bivariate logistic regression was used to identify associations between CGM metrics or glycemic markers with pregnancy outcomes, specifically preeclampsia, preterm birth, LGA, neonatal hypoglycemia, and NICU admission. Outcomes were chosen to reflect maternal complications (preeclampsia), neonatal hyperinsulinemic complications (LGA, neonatal hypoglycemia), and outcomes with particular relevance for health economic outcomes (preterm birth, NICU admission). We chose to include unadjusted models only because these closely reflect decision making in clinical practice where glycemic targets are used consistently and are not adjusted according to other patient characteristics.

To compare variables with different units, results are presented as standardized odds ratios (ORs) with 95% CIs. We used receiver operating characteristic (ROC) curves to compare the predictive ability of different metrics and glycemic markers individually.

Two hundred twenty-five women were enrolled in CONCEPTT and gave birth to liveborn infants. Of these participants, 157 gave at least one additional sample for laboratory testing of alternative markers of glycemia. Participants had a mean age of 32 years, had a BMI of 25.5 kg/m², and were predominantly of European or Mediterranean origin. Around 50% of women used insulin pump therapy. Overall, participants were similar to women who did not give an additional laboratory sample (n = 70), although they were less likely to have previous diabetes complications (Table 1). Rates of preeclampsia (13%), preterm delivery (40%), LGA (62%), neonatal hypoglycemia (27%), and NICU admission (35%) were similar in women who did and did not participate.

CGM metrics and laboratory markers of glycemia varied across gestation (Table 2) and were significantly correlated (Supplementary Table 2). Participants had an initial HbA_1c of 51 mmol/mol (6.9%) in the first trimester, which decreased to 46 mmol/mol (6.3%) at 24 weeks and slightly increased to 47 mmol/mol (6.4%) at 34 weeks’ gestation (Table 1). The corresponding CGM TIR (63–140 mg/dL [3.5–7.8 mmol/L]) was 52% in the first trimester, 50% at 24–25 weeks, and 64% at 34–35 weeks (Table 2).

Most CGM metrics were associated with one or more outcomes, including preeclampsia, preterm delivery, LGA, neonatal hypoglycemia, and NICU admission (Fig. 1 and Supplementary Table 3; standardized ORs). No CGM metrics in the first trimester or at 34 weeks were associated with preeclampsia, but at 24 weeks, associations were identified with CGM mean glucose, TIR (63–140 mg/dL [3.5–7.8 mmol/L]), time above range (TAR) (>140 mg/dL [>7.8 mmol/L]), and glucose SD. For preterm birth, mean glucose, TIR, and TAR showed associations in the first and second trimesters, but only TAR remained significant at 34 weeks’ gestation.

For LGA, CGM metrics showed consistent associations with TIR, TAR, and SD at all time points studied; mean glucose was also significant at 24 and 34 weeks. For neonatal hypoglycemia, mean glucose, TIR, and TAR showed associations at 24 and 34 weeks, but only TIR had an association in the first trimester. For NICU admission, mean glucose, TIR, TAR, and SD all showed associations at 24 weeks, but only TIR was associated in the first trimester. There were no CGM metrics that could predict NICU admission at 34 weeks. Overall, CGM TIR and TAR showed the most consistent associations with neonatal outcomes.

All laboratory markers of glycemia were associated with one or more outcomes. No laboratory markers were associated with preeclampsia in the first trimester, but we identified associations between preeclampsia and glycated albumin or gCD59 at 24 weeks and fructosamine or 1,5-AG at 34 weeks’ gestation. HbA_1c was not associated with preeclampsia but was associated with preterm birth at 24 and 34 weeks’ gestation. Preterm birth was also associated with 1,5-AG concentrations at 12 weeks and gCD59 concentrations at 24 weeks’ gestation.

All laboratory markers of glycemia were associated with LGA at one or more of the time points studied. 1,5-AG in the first trimester and gCD59 at 34 weeks’ gestation demonstrated the strongest associations with LGA. Neonatal hypoglycemia was associated with gCD59 at all three time points, with HbA_1c at 24 and 34 weeks and with fructosamine and 1,5-AG at 24 weeks’ gestation. NICU admission was associated with gCD59 at 24 weeks and HbA_1c at 24 and 34 weeks’ gestation. gCD59 showed the strongest associations with neonatal hypoglycemia and NICU admission at 24 weeks’ gestation.

The direction of most significant associations was toward a higher risk of complications with increased maternal hyperglycemia. The only exceptions were for 1,5-AG with preterm birth in the first trimester and fructosamine and 1,5-AG with preeclampsia at 34 weeks’ gestation.

ROC curves were used to compare the ability of the laboratory markers of glycemia with CGM metrics (mean glucose, TIR, and TAR) to predict pregnancy outcomes (Fig. 2, only including the strongest CGM metrics, and Supplementary Table 4). However, as expected for interrelated glycemic markers, CIs for ORs and area under the ROC curve (AUROC) were often overlapping (Supplementary Tables 3 and 4). The strongest predictor (defined as having the highest AUROC) for preeclampsia was mean CGM glucose in the first trimester (AUROC 0.65), mean CGM glucose at 24 weeks (AUROC 0.72), and fructosamine at 34 weeks’ gestation (AUROC 0.76). For preterm birth, mean CGM glucose and TAR were equally predictive in the first trimester (AUROC 0.61 for both), while at 24 weeks, mean CGM glucose, TAR, and gCD59 were equally predictive (AUROC 0.64). HbA_1c was the strongest predictor of preterm birth at 34 weeks (AUROC 0.65).

For LGA, 1,5-AG and TIR were the strongest predictors in the first trimester (AUROC 0.64 for both); TIR, fructosamine, and HbA_1c at 24 weeks (AUROC 0.64 for each); and TAR at 34 weeks (AUROC 0.67). The strongest predictors of neonatal hypoglycemia were gCD59 in the first trimester and 24 weeks (AUROC 0.61 and 0.72, respectively) and HbA_1c at 34 weeks (AUROC 0.68). There was no significant predictor for NICU admission in the first trimester; gCD59 was the strongest predictor at 24 weeks (AUROC 0.73) and HbA_1c at 34 weeks’ gestation (AUROC 0.66).

HbA_1c, CGM metrics, and alternative laboratory markers of glycemia can all be used to identify pregnancies at increased risk of suboptimal neonatal outcomes, even from the first trimester. However, neither laboratory markers nor CGM metrics were able to provide a strong prediction of any pregnancy outcome (AUROCs mostly <0.70). In pregnant women with type 1 diabetes, the use of alternative laboratory markers did not appreciably increase the AUROC for prediction of suboptimal pregnancy outcomes beyond HbA_1c, which is already widely available, or CGM metrics, such as TIR and TAR.

HbA_1c was consistently associated with pregnancy outcomes, suggesting that despite the known limitations of HbA_1c for assessing antenatal glycemia (14), it is still a critically important biomarker for obstetric and neonatal health outcomes. While other laboratory biomarkers demonstrated some promise, none were able to significantly increase the AUROC, showing, at best, comparable prediction to HbA_1c alone.

Glycated albumin and fructosamine (measuring total glycated serum proteins) provide a measure of glycemic status over the previous 2–4 weeks and have been associated with the development of diabetes and microvascular complications in nonpregnant adults (15,16). Although fructosamine and glycated albumin are unaffected by altered hemoglobin concentration, potential disadvantages to the use of fructosamine include the variability of albumin and protein concentration as a result of the dilutional effects of pregnancy so that fructosamine falls with advancing gestational age (17). Glycated albumin, expressed as a proportion of total albumin concentrations, shows less variation with gestational age and iron deficiency (14).

After initial interest in the use of glycated proteins for gestational diabetes mellitus (GDM) screening (18) and later dismissal because of low sensitivity (19), few studies have assessed their relationship with pregnancy outcomes. A study in 301 GDM and pre-GDM pregnancies showed that fructosamine concentration at delivery was not associated with neonatal outcomes (20). Glycated albumin may be more closely associated with HbA_1c and mean capillary blood glucose compared with fructosamine (21). Despite these findings, glycated albumin did not show strong associations with pregnancy outcomes in this study; even when outside pregnancy, it is associated with diabetes complications (22).

Another potential glycemic marker is 1,5-AG, a monosaccharide present in many foods that enters the body orally and is excreted gradually by the kidneys. In periods of hyperglycemia, glucose competes with reuptake in the renal tubule, promoting excessive loss of 1,5-AG. Thus, 1,5-AG falls during hyperglycemia and takes 2 weeks to restabilize once normoglycemia is restored (23). 1,5-AG has been found to be inversely associated with birth weight and with neonatal hypoglycemia in GDM (7,24). Nowak et al. (25) identified that 1,5-AG was associated with mean glucose and a strong predictor of macrosomia in 58 women with type 1 diabetes. Our results are consistent with these findings, as 1,5-AG was negatively associated with LGA in each trimester but did not perform better than HbA_1c values. A strong positive association between 1,5-AG and preeclampsia was observed at 34 weeks’ gestation, which was accompanied by a strong negative association of fructosamine and preeclampsia. Our observation of “improved” glycemic status at 34–35 weeks in women with preeclampsia (either already concurrent or to emerge within the next weeks) could indicate early changes in renal function or placental dysfunction with ensuing lower insulin requirements (reverse causation) (26). It is possible that these findings could be used to develop a predictive model in the future.

CD59 is a membrane protein inhibitor of the terminal complement cascade. Glycation of CD59 (yielding gCD59) abrogates its function as an inhibitor of complement, which likely plays a role in the pathogenesis of diabetes complications (27,28). gCD59 has been proposed as a novel marker of glycemic control that reflects changes within 2 weeks of a treatment intensification (6). In a study of 1,000 pregnant women, Ghosh et al. (6) identified that gCD59 is a strong predictor of GDM (AUROC 0.92) and that higher levels of maternal gCD59 are associated with a higher prevalence of LGA. More recently, measurement of gCD59 at pregnancy week <20 identified early development of GDM and was strongly associated with LGA newborns (29). In the current study, gCD59 measured at 24 weeks showed strong associations with neonatal hypoglycemia and NICU admission (AUROC 0.72–0.73) and performed better than HbA_1c (AUROC 0.64–0.66) at this time point. This suggests that if it was more widely available, gCD59 could potentially play a role in prediction of neonatal complications, particularly NICU admission and neonatal hypoglycemia.

Relatively few studies have assessed the predictive ability of CGM metrics during pregnancy (30). Mulla et al. (31) found no associations between HbA_1c or CGM metrics and birth weight or LGA in 41 pregnant women with type 1 diabetes. Likewise, Panyakat et al. (32) found no associations between third trimester CGM metrics and pregnancy outcomes in 47; women with GDM. Dalfrà et al. (33) reported that CGM glycemic variability indexes were associated with ponderal index in the newborn offspring of 32 women with type 1 diabetes but not in GDM. Law et al. (34) identified an association between nocturnal hyperglycemia and LGA but with no associations identified with other CGM metrics, such as TIR or TAR, in 162 women with GDM. The same team identified associations between LGA and lower mean CGM glucose during the first trimester and a higher mean CGM glucose during the second and third trimesters in women with type 1 or type 2 diabetes (35). Previous data from CONCEPTT demonstrated particular diurnal periods, reflecting postmeal hyperglycemia, when mean CGM glucose is higher in women who deliver an LGA infant (11). In this study, TAR was the CGM metric that showed the strongest association with pregnancy outcomes followed by TIR, which is the metric most commonly used in diabetes clinics because of its association with microvascular outcomes (36) and intuitive use both for patients and health care professionals (37). Overall, these results indicate that TIR and TAR, in addition to aiding daily self-management of diabetes, can also offer insight into the pregnancy outcomes of women with type 1 diabetes, predictions that may be further improved with new-generation CGM sensors yielding accurate CGM metrics for longer time periods throughout pregnancy. In regular CGM users, additional biomarkers, including HbA_1c, may not add to the assessment of maternal glycemia but remain important for prediction of pregnancy outcomes.

These results also provide insight regarding relevant time windows for pregnancy outcomes. Essentially, most associations, including preeclampsia, were more prominent in the second trimester, highlighting the importance of glycemic status in this period. Preeclampsia is generally agreed to commence in early pregnancy, but the current results are in line with other reports that also describe the association with maternal hyperglycemia as being only present or more marked in the second and third trimesters (38,39).

This study provides detailed information about laboratory markers and CGM metrics in a well-characterized cohort during each trimester of pregnancy. The study design provided a near-complete data set, albeit with fewer third trimester data. In addition, we had a robust process for analysis of additional glycemic markers, with batch analysis for consistency. We also acknowledge the limitations. Women with HbA_1c <6.5% (48 mmol/mol) or >10.0% (86 mmol/mol) at baseline were excluded from CONCEPTT, which may have reduced the strength of association between glycemic markers and pregnancy outcomes. CGM metrics and glycemic biomarkers were only measured at three time points, so it is possible that more frequent assessment, or use of CGM for >6 days duration and preferably continuously throughout pregnancy, may give further information and closer associations with obstetric and neonatal outcomes. Indeed, the optimal frequency of monitoring of CGM or laboratory biomarkers in pregnancy is unknown. This is particularly important for the novel laboratory markers, which have not been fully characterized in pregnant women and may show more dynamic changes than HbA_1c. While pregnancy outcomes were adjudicated, the timing of delivery was determined by local policies, and we cannot exclude variation in local criteria used for NICU admission. The 34-week sampling time was chosen to include almost all women, but several deliveries occurred before 34 weeks, which may have also affected our results. We chose not to adjust for multiple testing in this analysis for several reasons. First, adjustment of the threshold of significance is challenging where there are multiple interrelated variables and the options available do not facilitate a clear presentation of the data. Furthermore, because the analysis was primarily assessing the same question (associations between glycemia and outcomes), adjustment for multiple testing became less valid. We expected to find multiple significant associations and consider that our results are in line with the substantial body of evidence describing associations between maternal glucose and complications related to fetal hyperinsulinemia in type 1 diabetes pregnancy. We demonstrate that although some markers appeared stronger than others, there was a high degree of overlap of CIs for prediction of pregnancy outcomes. This precludes a definitive conclusion about which single marker performed best but is not unexpected since so many markers showed a high degree of statistical correlation.

In conclusion, this study provides a comprehensive assessment of measures of glycemia, using both CGM and laboratory markers to predict a range of outcomes in type 1 diabetes pregnancy. Despite the established importance of glycemia in type 1 diabetes pregnancy (9,40), markers of glycemia were only moderate predictors of outcomes (AUROC mostly <0.70). It is possible that the complexity of maternal hyperglycemia cannot be easily summarized by a single glycemic marker, or that other maternal, fetal, and placental factors also contribute directly or indirectly, by mediating the relationship between glycemia and pregnancy outcomes. This is consistent with the finding that complications such as LGA have not substantially improved over recent decades, despite advances in diabetes management and technology (1,9,31). Future work should seek to optimize maternal glucose levels before and during pregnancy and to identify what factors in addition to maternal hyperglycemia contribute to the variation in outcomes seen in type 1 diabetes pregnancy.

Acknowledgments. The authors thank all the women with type 1 diabetes who participated. The authors also acknowledge the invaluable support from the 31 clinical care teams and the CONCEPTT Steering Committee: Denice S. Feig, Helen R. Murphy, Elisabeth Asztalos, Jon F.R. Barrett, Rosa Corcoy, Alberto de Leiva, Lois E. Donovan, J. Moshe Hod, Lois Jovanovic, Erin Keely, Craig Kollman, Ruth McManus, Kellie E. Murphy, Katrina Ruedy, Marlon Pragnell, Olivia Lou, Aaron Kowlaski, and George Tomlinson.

Dr. Lois Jovanovic died during the preparation of the manuscript.

Funding. The trial is funded by JDRF grant 17‐2011‐533 and grants under the JDRF Canadian Clinical Trial Network, a public‐private partnership including JDRF and FedDev Ontario and supported by JDRF 80‐2010‐585. Medtronic supplied the CGM sensors and CGM systems at reduced cost. This ancillary project was funded by the European Foundation for the Study of Diabetes (EFSD)/Sanofi Pilot Research Grants for Innovative Measurement of Diabetes Outcomes, 2017. Reagents for analysis of glycated albumin were provided by Asahi Kasei Pharma through Spinreact. Reagents for 1,5-AG were provided by Hirotaka Ishibashi at GlycoMark. Laboratory analysis for these analytes was performed in the National Institute for Health Research Core Biochemistry Assay Laboratory, Cambridge Biomedical Research Campus. Reagents and laboratory analysis for gCD59 were provided by J.A.H. C.L.M. is supported by the Diabetes UK Harry Keen Intermediate Clinical Fellowship (DUK-HKF 17/0005712) and the EFSD-Novo Nordisk Foundation Future Leaders Award (NNF19SA058974). H.R.M. conducts independent research supported by the National Institute for Health Research (Career Development Fellowship CDF-2013-06-035) and is supported by Tommy’s charity.

Duality of Interest. D.S.F. has received honoraria for speaking engagements from Medtronic and has been on an advisory board for Novo Nordisk. E.M.S. has received honoraria for speaking engagements with Eli Lilly and Abbott Diabetes Care and has been on advisory boards for Abbott Diabetes Care. J.A.H. has a financial interest in Mellitus LLC. Mellitus has licensed intellectual property for the technology used in this research and in developing diagnostic tools for diabetes. The interests of J.A.H. are reviewed and managed by Brigham and Women’s Hospital and Partners HealthCare in accordance with their conflict of interest policies. H.R.M. has received honoraria for speaking engagements from Medtronic, Roche, Novo Nordisk, and Eli Lilly and is a member of the Medtronic European Advisory Board. R.C. has received honoraria for speaking engagements with Eli Lilly and Novo Nordisk and has been on advisory boards for Novo Nordisk and Abbott. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.L.M. designed the study; arranged the laboratory analysis for 1,5-AG, fructosamine, and glycated albumin; collated, analyzed, and interpreted the data; and wrote and revised the manuscript. D.T. contributed to the analysis and interpretation of the data. D.S.F. and H.R.M. identified the study question, contributed to data analysis and discussion, and reviewed and revised the manuscript. J.M.Y. contributed to quality control of the database. E.M.S. provided expertise on analysis of CGM data. D.D.M. and J.A.H. analyzed the samples for gCD59 and contributed to the discussion. R.C. identified the study question, designed the study, contributed to data analysis and discussion, and reviewed and revised the manuscript. All authors reviewed the final version of the manuscript before publication. C.L.M. is the guarantor of this work and, as such, had 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.