The prevalence of gestational diabetes mellitus (GDM) has been increasing in Australia and worldwide likely due to rising average maternal age and increasing obesity especially in young adults (1). In Australia, GDM is now becoming a common complication of pregnancy as it affects around 10% of pregnancies with up to 30% of pregnancies being affected by GDM in high-risk populations (2). In most cases, GDM occurs in pregnant women with impaired pancreatic function, which is insufficient to overcome the insulin resistance associated with the pregnant state.

In recent years, GDM has been diagnosed more frequently in Australia based on more stringent diagnostic criteria recommended by the International Association of Diabetes and Pregnancy Study Groups (IADPSG), which were endorsed by the Australasian Diabetes in Pregnancy Society (ADIPS) (2). The usual time point for GDM screening is recommended to be between 24-28 weeks of pregnancy (3). Earlier screening, recommended by these expert groups, in high risk women is desirable to enable lifestyle interventions focused on diet, physical activity, and weight control to be initiated during the first or early second trimesters of pregnancy (4).

The diagnosis of GDM carries important risks of adverse short and long-term clinical outcomes for women and their offspring. The main immediate consequences of GDM are increased risks of preeclampsia, large for gestational age (LGA) newborns, and caesarean birth, with their associated perinatal co-morbidities (5). GDM is associated with up to 10-times higher odds for the development of future maternal type 2 diabetes or prediabetes in comparison with individuals with a normoglycemic pregnancy (6). Women who are affected by GDM are not only at high risk of developing type 2 diabetes later in life, furthermore, having gestational diabetes is associated with a relative risk of 2.0 (95% CI, 1.6-2.5) for being affected by future cardiovascular disease (7).

Importantly, pharmacological treatment has been shown to improve perinatal outcomes of GDM with reductions in preeclampsia, macrosomia, shoulder dystocia and neonatal death (8). Insulin has been recommended as the first-line treatment agent for GDM in the U.S (9) while in the UK, the National Institute for Health and Care Excellence (NICE) together with Scottish and Canadian guidelines recommends that metformin, an insulin sensitizer, which reduces hepatic gluconeogenesis, and increases peripheral glucose uptake (10), may be considered as initial pharmacological glucose lowering treatment in GDM women (11). Although insulin therapy has been shown to reduce the risk of neonatal macrosomia and rate of serious perinatal outcomes such us shoulder dystocia or perinatal death (12), the benefits of insulin treatment in pregnancy often do not extend to preventing neonatal hypoglycaemia, frequently requiring intravenous glucose infusion and neonatal high-level nursery admission (13). Furthermore, gestational insulin therapy requires additional education, resources and training with the need for increased care for women throughout pregnancy and the act of injecting insulin can be stressful for some women.

At our institution, in our cohort of pregnant women with GDM, metformin has been endorsed as an alternative treatment to insulin therapy. We have therefore hypothesized that metformin use to treat GDM will result in similar pregnancy outcomes in comparison to pregnant women who are treated with insulin alone.

The present study examined perinatal outcomes for women with GDM who were treated with pharmacological interventions in comparison with dietary lifestyle changes alone while controlling for differences in baseline maternal characteristics. In particular, the primary aim was to examine the differences in composite maternal and in neonatal outcomes between four GDM treatment groups (“Diet”, “Metformin”, “Metformin and Insulin”, “Insulin”). In addition, specific maternal and neonatal outcomes were also reported and analysed separately. The second aim of the study was to identify early clinical maternal predictors for the use of pharmacological treatment in pregnancy affected by GDM.

The results from the present study support our hypothesis that metformin treatment (alone or combined with insulin) of women with GDM does not result in worse pregnancy outcomes as compared with those who were assigned to insulin. Despite no significant difference in main composite study aim between study groups and Diet group, insulin treatment alone, although prescribed for small number of GDM patients, was associated with a higher proportion of “small for gestational age” neonates and higher rates of neonatal hypoglycaemia in comparison with dietary or metformin treated groups. Such neonatal complications likely resulted from worse glycaemic control in this group of patients as indicated by their high daily maternal insulin requirements, which ranged between 58 units to 104 units.

Importantly, previous large retrospective cohort study reported that pregnancy outcomes are worse in GDM women with SGA neonates than in those without GDM with subsequent increased risks for respiratory distress syndrome, intrauterine foetal death, hypoglycaemia, jaundice and neonatal demise (19).Interestingly however, the metformin treated group in comparison with dietary intervention group had a higher risk for LSCS. Such association between mode of birth and metformin treatment is related not only to the treatment but also to the differences in our subjects basal characteristic. Previous studies have highlighted that, increased maternal BMI either in overweight or obese category without GDM, increased the risk of macrosomia and caesarean delivery when compared to normal weight women (20). In our metformin treated cohort we observed no evidence of foetal growth acceleration on third trimester ultrasound in majority of patients, likely reflective of satisfactory maternal glucose control (21). Indeed, once we excluded elective LSCS from the analysis there was no difference in number of emergency LSCS in metformin group in comparison with dietary intervention.

To our knowledge no randomised trials (RCTs) compared effects of metformin directly to dietary/lifestyle intervention in pregnancy, although several previous studies compared metformin and insulin interventions alongside dietary interventions for both trial arms. The analysis of 16 RCTs or follow-up of a RCTs revealed that metformin in comparison with insulin treatment did not increase the risk of caesarean section (RR = 0.97; 95% CI, 0.80 to 1.19) (22). Furthermore, the meta-analysis consisting of 11 trials reported that women randomised to metformin had lower risk for adverse maternal and neonatal outcomes including lower risk for the instrumental delivery compared to those randomised to insulin (23). The above data points to metformin being a useful alternative to insulin therapy with a high degree of patient acceptability.

Importantly, in comparison with insulin, metformin can significantly decrease maternal weight gain, and therefore metformin is now the preferred treatment option for an increasing number of women with a BMI in obese category (24). The efficacy of metformin treatment in GDM is not without limitation, as congruent with present study, approximately 14% to 46% of pregnant women fail to achieve adequate glycaemic control with metformin alone (25).

The lack of longer-term safety studies and that metformin can cross the placenta raise potential concerns associated with metformin therapy in pregnancy. The safety and optimal metformin doses in pregnancy have not been yet defined, however most studies use metformin doses ranging from 500 mg to 2500 mg a day (22). In the present study the metformin treatment was commenced in later stages of pregnancy, on average at 28 ± 6.3 weeks of pregnancy, therefore without effect on early embryonic growth.

To date, no increased risk for non-genetic congenital anomalies has been identified following foetal exposure to metformin during the first trimester of pregnancy. A randomized, placebo-controlled trial of PCOS women who were either randomised to metformin (500 mg twice daily increasing to 1000 mg twice daily) or placebo from the first trimester gestational age between 5 and 12 weeks found no difference in the primary composite study outcome of preeclampsia, GDM and preterm delivery (26). Furthermore, no adverse safety signal was detected in randomised, placebo controlled trial of pregnant women with type 2 diabetes in pregnancy who were randomised either to metformin or placebo at 16.5 weeks of pregnancy. This study found no significant difference in congenital anomaly with 7/227 (3%) affected infants in metformin treated group in in comparison with 13/227 (6%) infants in dietary intervention group (p value of 0.16, RR 0.52 (0.22 to 1.28)) (27). Reassuringly, in previous studies, exposure in utero in children of GDM women to metformin ( ± insulin) or insulin alone led to similar total and abdominal body fat percent and metabolic measures at children at 7–9 years of life (28). Conversely, in a recent study, metformin exposure in the first trimester of pre-gestational diabetes was associated with an increased risk of birth defects and pregnancy loss; however, these adverse pregnancy outcomes were attributed to underlying disease rather than to metformin therapy (29). On-going long term follow-up studies of children born to mothers affected by GDM will help answer this current uncertainty.

In the present study, only 33% of GDM patients achieved satisfactory glycaemic control through the dietary therapy while the majority of pregnant GDM women (77%) required pharmacological treatment. The ability to predict a-priori which GDM group of patients will fail their dietary intervention would help to plan steps more effectively in their GDM management.

Past studies have identified the following predictors of the requirement to introduce pharmacological treatment in GDM. These predictors included early GDM diagnosis (e.g., at <25 weeks gestation), a family history of diabetes, non-European ethnicity, an older age, elevated fasting blood glucose level, HbA1c at GDM diagnosis, and an elevated pre- pre-pregnancy BMI (20, 30). In our unique group of pregnant non-obese women, we have identified fewer predictors of the need for pharmacological GDM treatment. Those predictors included maternal characteristics such as baseline maternal BMI, value of fasting BGL on their OGTT, the number of previous miscarriages and early GDM diagnosis (e.g., at <24 weeks gestation). Although on average Caucasian women had a higher fasting glucose on their GTT than women of Asian ethnicity, their ethnicity was not a significant predictor of the need for escalating GDM therapy beyond diet alone. Maternal BMI, once controlled for the fasting BGL level on OGTT and timing of the OGTT and number of previous miscarriages, was no longer a significant indicator of the need for pharmacological GDM therapy.

In previous research, elevated pre-pregnancy maternal BMI predicted failure of dietary therapy (20) with higher maternal BMI being significantly associated with the need for medical treatment (31–33). Although obesity is associated with increasing insulin resistance and pancreatic β-cell dysfunction, it remains unclear whether weight control during pregnancy, as recommended by the Institute of Medicine, would reduce the risk of GDM or the need for insulin therapy (34). Considering that our study patients had an average BMI close to the normal range at 25.8 ± 4.7 kg/m², we may hypothesise that having an initial BMI in the obese category would have been more closely associated with the need for pharmacological GDM intervention.

In the present research, the value of fasting glucose level on the OGTT was the strongest indicator that women with GDM may not respond to the dietary intervention alone. This study finding is important considering that some at risk women are unable to complete OGTT in pregnancy. Interestingly we have also found the increased risk for LSCS and instrumental vaginal birth with raised fasting BGL in metformin treated group. Interestingly, once the effect of metformin treatment on mode of birth was controlled for fasting glucose on subjects’ OGTT and subjects initial BMI, the risk for LSCS or instrumental vaginal birth increased significantly in metformin treated GDM women. The previous retrospective cohort study of 14,741 pregnant women found that fasting hyperglycaemia was associated with increased risk for caesarean birth (OR: 1.33, 95% CI: 1.15-1.55,P < 0.001) (35). Such strong positive association between fasting hyperglycaemia on OGTT and adverse perinatal outcomes, including caesarean birth, was noted in previous systematic review and meta-analysis of GDM women (36).

Multiple studies have highlighted the link between fasting hyperglycaemia in the first and 2^nd trimester of GDM pregnancies with increased occurrence of adverse pregnancy outcomes including the need for surgical birth (4, 37, 38). In previous study only fasting plasma glucose value on the oral glucose tolerance test in pregnancy was significantly associated with pregnancy adverse outcomes, irrespectively of pharmacological intervention (39). Therefore, effective treatment of fasting BGL in women affected by GDM may potentially improve maternal and neonatal health outcomes with a great potential for the early detection of women at risk of having more adverse perinatal outcomes, irrespective of their other risk factors, such as obesity and maternal age.

Interestingly, we observed that having at least one previous miscarriage or pregnancy termination may influence the need for pharmacological treatment for GDM women. Several murine studies have reported that progesterone, which is essential to sustain pregnancy, promotes insulin resistance by multiple mechanisms during pregnancy (40–42). Interestingly, previous case control study of 1567 Korean women demonstrated that threatened miscarriage is associated with decreased risk of GDM and the severity of glucose intolerance (43). Conversely, a retrospective cohort study found that having a spontaneous miscarriage was linked to a higher risk for having subsequent gestational diabetes (44). Further research is recommended to confirm these relationships and to evaluate the pathophysiologic mechanisms that interplay between these common obstetric complications.

Several studies have shown that vitamin B12 status during pregnancy is important to the health of mother and her offspring (45). In the present study, limited by the small sample size, approximately 10% of women had total vitamin B12 level in the insufficiency range and their vitamin B12 levels declined significantly during pregnancy. As metformin treatment may reduce ileal absorption of dietary vitamin B12 (46), GDM women treated with metformin would likely benefit from monitoring of their vitamin B12 status.

Our study is not free of limitations, due to pragmatic method of data collection as part of subjects’ routine clinical care, rather than at fixed short time intervals. Therefore detailed trajectories of weight gain and glycaemic control during pregnancies were not analysed in this study. However an absence of foetal growth acceleration on third trimester ultrasound in the majority of patients was likely reflective of their satisfactory maternal glucose control (21). Exclusion from the analysis of severely obese women with BMI> 35 kg/m² might have reduced the risk of their pregnancy complications and a rate of instrumental or caesarean birth. Due to our small number of adverse perinatal events, and in order to improve the ability to detect differences in the primary study endpoints as well as to increase study statistical power, we have designed the main composite study outcome of combined maternal and neonatal events. Additionally, we have reported detailed information on clinically important events of which the composite outcome is based on, with their measures of association ( Table 2 ). Additional advantage of this cohort study is consistent and uniform nutritional counselling as well as consistency of care being provided by the same physician.

Our clinical cohort study was relatively small. It is possible that with a bigger cohort, the association between pharmacological interventions and adverse perinatal outcomes may reach statistical significance. However, for the second study aim, when we grouped all pharmacological interventions together, we were able to identify significant predictors for the need of pharmacological treatment in GDM.

In summary, our study addressed the paucity of existing data comparing the effect of metformin intervention in pregnancy with dietary/lifestyle intervention in women with BMI below 35 kg/m² and gestational diabetes. The present study has highlighted that metformin treatment of GDM may not be associated with different pregnancy outcomes compared to the GDM managed by diet except for the increased risk for the LSCS. These study findings, however, were no longer significant once the analysis was controlled for the number of elective caesarean sections.

We have also observed that elevated fasting blood glucose on OGTT, in metformin treated GDM women, is a stronger predictor of their need for either instrumental delivery or caesarean section. Moreover, due to combined demographic, obstetric and medical data we have identified the local characteristics of women with GDM, which would help to predict their need for pharmacological therapy. This predictive model will improve streamlining of our patients’ care and improve utilization of local hospital resources. Further studies are still needed to identify the most effective and safe management of gestational diabetes within the public hospital setting.

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by The Southern Eastern Sydney Local Health District Human Research Ethics Committee (Study Reference No. RESP/15/107). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Study design: MMB, AP, DB, AO'S. Study conduct: MMB. Data collection: MMB, AP. Data analysis: MMB, AP, DB, AKP. Data interpretation: MMB, AP, DB, AKP, AZ, AO'S. Drafting manuscript: MMB. Revising manuscript content: MMB, AP, DB, AKP, AZ, AO’S. Approving final version of manuscript: all authors. MB is the guarantor of this work. All authors contributed to the article and approved the submitted version.