Short-term neonatal outcomes in women with gestational diabetes treated using metformin versus insulin: a systematic review and meta-analysis of randomized controlled trials

Aims To expand the evidence base for the clinical use of metformin, we conducted a meta-analysis of randomized controlled trials (RCTs) comparing the efficacy and safety of metformin versus insulin with respect to short-term neonatal outcomes. Methods A comprehensive search of electronic databases (PubMed, Embase, Cochrane Library, and Web of Science) was performed. Two reviewers extracted the data and calculated pooled estimates by use of a random-effects model. In total, 24 studies involving 4355 participants met the eligibility criteria and were included in the quantitative analyses. Results Unlike insulin, metformin lowered neonatal birth weights (mean difference − 122.76 g; 95% confidence interval [CI] − 178.31, − 67.21; p < 0.0001), the risk of macrosomia (risk ratio [RR] 0.68; 95% CI 0.54, 0.86; p = 0.001), the incidence of neonatal intensive care unit admission (RR 0.73; 95% CI 0.61, 0.88; p = 0.0009), and the incidence of neonatal hypoglycemia (RR 0.65; 95% CI 0.52, 0.81; p = 0.0001). Subgroup analysis based on the maximum daily oral dose of metformin indicated that metformin-induced neonatal birth weight loss was independent of the oral dose. Conclusions Our meta-analysis provides further evidence that metformin is a safe oral antihyperglycemic drug and has some benefits over insulin when used for the treatment of gestational diabetes, without an increased risk of short-term neonatal adverse outcomes. Metformin may be particularly useful in women with gestational diabetes at high risk for neonatal hypoglycemia, women who want to limit maternal and fetal weight gain, and women with an inability to afford or use insulin safely. Supplementary Information The online version contains supplementary material available at 10.1007/s00592-022-02016-5.


Introduction
Gestational diabetes mellitus (GDM) is a common complication during pregnancy and is defined as any glucose intolerance that occurs or is diagnosed for the first time during pregnancy [1]. GDM develops in about 5% to 14% of all 1 3 pregnancies and is associated with certain pregnancy-related complications and a long-term risk of diabetes in both the mother and offspring [2]. With the establishment of the twochild policy and epidemic of obesity in China, the incidence of GDM has been increasing, resulting in a heavy economic burden on the public health care system and individuals [3]. According to the latest data reported by the International Diabetes Federation in 2021, about one in six live births (20 million) is affected by high plasma glucose concentration during pregnancy, and GDM accounts for 83.6% of these cases of hyperglycemia [4].
Women with uncontrolled GDM have higher-risk pregnancies, and some adverse effects of GDM may also affect the fetus, including fetal anomalies, macrosomia (birth weight of > 4000 g), fetal distress, metabolic disorders, growth imbalance, hyperbilirubinemia, and some long-term complications [5]. Traditionally, insulin has been the gold standard for the treatment of GDM because it cannot cross the placenta and allows for precise glucose control. However, insulin therapy has several disadvantages, including the need for multiple injections, risks of hypoglycemia and hyperbilirubinemia, the rising cost of insulin, and the lack of affordability [6]. These disadvantages suggest that current treatment regimens fall short of optimizing outcomes. Metformin is a commonly used oral antihyperglycemic drug in clinical practice with excellent efficacy in terms of glycemic control and weight loss, good tolerance, and a reasonable price [7]. Several organizations currently support its use as an alternative to insulin [8,9]. However, recent long-term studies of offspring have provided conflicting results. Two follow-up studies of children aged 2 to 9 years whose mothers had gestational diabetes showed that several growth parameters tended to be larger in metforminexposed offspring than in offspring exposed to insulin. These growth parameters included weight, body mass index, triceps skinfold, waist and arm circumferences and body fat percent, and they were also associated with cardio-metabolic disease in later life [10,11]. This has slowed the clinical use of metformin as a substitute for insulin in the treatment of GDM.
We therefore performed this updated meta-analysis to compare the efficacy and safety of metformin versus insulin with respect to short-term neonatal outcomes in the treatment of GDM. The objective of our study was to determine whether metformin is superior to insulin in terms of altering neonatal growth outcomes and inducing neonatal adverse outcomes during treatment of GDM. Addressing this issue is particularly important because the number of pregnancies exposed to metformin is increasing worldwide.

Methods
This systematic review and meta-analysis are reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement and was registered at the International Prospective Register of Systematic Reviews (CRD42022330187) [12].

Search strategy
A systematic literature search of PubMed, Embase, the Cochrane Library, and Web of Science (last search was updated on 1 May 2022) was performed using prespecified terms (Supplemental Text S1) with no filters and no language or location restrictions. We also searched for additional eligible trials in previously published meta-analyses on related topics.

Inclusion and exclusion criteria
Studies that met the following criteria were included: (1) The population comprised pregnant women with GDM, (2) The interventions were metformin (with or without extra insulin treatment) and insulin, (3) The study included one or more neonatal outcomes, and (4) The study design was a randomized controlled trial (RCT). We excluded studies involving pregnant women with pre-existing diabetes, and duplicate studies published in different journals were included only once.

Definitions of neonatal outcomes
The neonatal outcomes included neonatal growth outcomes and neonatal adverse outcomes. The neonatal growth outcomes were birth weight, birth height, macrosomia (≥ 4000 g), large for gestational age (LGA) (birth weight at the > 90th percentile), and small for gestational age (SGA) (birth weight at the < 10th percentile). The neonatal adverse outcomes were neonatal hypoglycemia, admission to the neonatal intensive care unit (NICU), hyperbilirubinemia, respiratory distress syndrome, premature birth, congenital anomalies, abnormal pH of the umbilical cord, abnormal Apgar score at 5 min, neonatal death, neonatal sepsis, and birth trauma.

Data collection and management
The titles, abstracts, citation information, and descriptor terms of the publications identified through the search strategy were screened. Full-text articles of all selected abstracts were obtained, and two reviewers (Bo Sheng and Juan Ni) independently assessed all the full-text articles for eligibility to determine the final study selection. Any disagreements between the two authors were settled by group discussion until a consensus was reached. We designed a data extraction form to collect relevant information including the authors, year of publication, country, number of patients, definition of gestational diabetes, patient characteristics, and interventions.

Risk of bias and quality assessment
We used the Cochrane Collaboration's tool to assess the risk of bias in terms of the following seven aspects: (1) Random sequence generation (selection bias), (2) Allocation concealment (selection bias), (3) Blinding of participants and personnel (performance bias), (4) Blinding of outcome assessment (detection bias), (5) Incomplete outcome data (attrition bias), (6) Selective reporting (reporting bias), and (7) Other bias. We classified these aspects as low risk of bias, uncertain risk of bias, or high risk of bias.
We assessed the quality of evidence in these studies by using the GRADE profiler (GRADEpro GDT) [13]. The GRADE system was used to assess the study limitations (risk of bias), inconsistency, indirectness, imprecision, and publication bias across the body of evidence to derive an overall summary of the quality of evidence, which was classified each as high, moderate, low, or very low.

Statistical analysis
The standardized mean difference (SMD) was calculated using the mean and standard deviation for continuous variables. The risk ratio (RR) was calculated for dichotomous variables with 95% confidence intervals (CIs). The metaanalysis was performed using Review Manager (RevMan) version 5.4.1 (Nordic Cochrane Centre, Copenhagen, 2014), and Egger's test was used to assess publication bias through the 'metafor' package in R version 3.5.1 [14]. The studies were determined to be heterogenous if I 2 > 50% and p < 0.1. A sensitivity analysis was performed by excluding each study one by one to evaluate the credibility of the pooled results. A prespecified subgroup analysis was also performed to explore the sources of heterogeneity. Potential publication bias was assessed by the application of contour-enhanced funnel plots and Egger's linear regression test at the p < 0.05 level of significance. If publication bias was indicated, we further evaluated the number of missing studies by trimand-fill analysis and recalculated the pooled risk estimate with the addition of those missing studies. Except where otherwise specified, a p value of < 0.05 was considered statistically significant.

Literature search and study characteristics
In total, 576 studies were retrieved through PubMed, Embase, the Cochrane Library, and Web of Science. After removal of duplicates and title/abstract screening, 188 trials underwent full text assessment, after which the full set of eligibility criteria was applied. After full text evaluation, 24 studies remained eligible for inclusion in this review. The process of study selection is illustrated in Fig. 1. As shown in Table 1, 24 RCTs involving 4355 patients with GDM were included to estimate the impact of metformin versus insulin on neonatal outcomes . The earliest study began in 2001, and the latest study was completed in 2021. Five studies each were conducted in Iran [23,[30][31][32]35],Egypt [15,18,20,22,37], and Pakistan [16,17,21,27,38]; three in Finland [26,28,34]; and one each in Australia [19], India [36], Spain [29], Brazil [33], New Zealand [24], and the USA [25]. In this meta-analysis, we mainly focused on the daily oral dose of metformin in pregnant women with GDM. Three studies among the 24 RCTs did not report the dose of metformin [19,23,30], and the remaining 21 studies were included for further subgroup analysis.
Supplemental Fig. S1 provides a summary of the risk of bias for each included study. No selection bias, attrition bias, or selective bias was present in any of the RCTs, indicating relatively high quality. Because insulin was given by injection and metformin was given orally, all the included studies involved open allocation, which did not affect the short-term neonatal outcomes because these were all objective. The quality of the evidence (GRADE) for the neonatal outcomes of interest, including neonatal birth weight, macrosomia, LGA, SGA, birth height, NICU admission, and neonatal hypoglycemia, was very low to moderate. The GRADE system evidence for the above outcomes and reasons for upgrade and downgrade are shown in Table 2.
To assess the potential publication bias of neonatal birth weight, we used the 'metafor' package of R software for Egger's test. Our results showed that the funnel plot of neonatal birth weight was asymmetrical (Supplemental Fig.  S3A), and Egger's test indicated possible publication bias (p = 0.008) (Supplemental Table S1). Next, we used trimand-fill analysis to recalculate our pooled risk estimate; the results suggested no publication bias (p = 0.28), and the funnel plot also became symmetrical (Supplemental Fig. S3B).

Discussion
In this systematic review and meta-analysis, we found that neonates exposed to metformin in utero weighed less at birth than those whose mothers were exposed to insulin. The risk of macrosomia is substantially lower (by 30%) when GDM is treated with metformin than with insulin, LGA. Despite being born at lower average birth weights, neonates of metformin-treated women do not have an increased incidence of neonatal adverse outcomes. In contrast, metformin significantly lowers the risk of neonatal hypoglycemia and the incidence of NICU admission.
It is well accepted that the fetuses of obese women with GDM have a higher risk of developing macrosomia than those of women with GDM of normal weight [39]. Some recent meta-analyses showed that weight gain during pregnancy was significantly lower in women with GDM who received metformin than in those who received insulin [4,7]. Whether metformin-induced weight loss in women with GDM leads to a significant reduction in the incidence of fetal macrosomia remains unclear. Our results provide evidence that metformin can also effectively control neonatal birth weight and reduce the incidence of fetal macrosomia. In particular, there is growing evidence that macrosomia is likely to be associated with shoulder dystocia, brachial plexus injury, delayed motor development, and a higher risk of obesity or diabetes later in life [7,40]. Moderate neonatal birth weight control may effectively reduce and avoid some complications related to macrosomia, especially for pregnant women with GDM. To explore the relationship between neonatal birth weight and the oral dose of metformin, we performed a subgroup analysis of neonatal birth weight based on the maximum daily oral dose of metformin. We found that a maximum oral dosage of metformin of 1500, 2500, and 3000 mg/day was associated with neonatal birth weight loss, but there was no significant difference in an oral dosage of metformin of 2000 and 2250 mg/day. These results suggest that metformin-induced neonatal birth weight loss occurs independently of the oral dose of metformin. This is consistent with the previous finding that a low dosage of metformin (< 1000 mg/day), but not a high dosage, had significant efficacy for body mass index control or weight loss in adolescents [41].
Macrosomic fetuses in women with diabetes develop a unique pattern of overgrowth involving central deposition of subcutaneous fat in the abdominal and interscapular areas with skeletal growth remaining largely unaffected [40,42].
During early gestation, the embryo expresses very low levels of organic cation transporters, making metformin likely to be safe in the first trimester. However, metformin can easily cross the placenta via organic cation transporters in the second and third trimesters and may reach near-maternal concentrations in the fetus [43]. In addition to lowering blood glucose concentration, metformin has a variety of intracellular effects including inhibition of mitochondrial respiration and effects on the nutrient-sensing pathway by both adenosine monophosphate-activated protein kinase and mammalian target of rapamycin mechanisms [44][45][46][47]. Moreover, in the Metformin in Women with Type 2 Diabetes in Pregnancy (MiTy) trial, the lower neonatal adiposity in the metformin group led to a lower incidence of fetal macrosomia [48]. Therefore, the significant metformin-induced reduction in the incidence of macrosomia may be related to the inhibition of fetal fatty acid synthesis. This effect of metformin differentiates its dose-dependent hypoglycemic effect, the underlying mechanism of which remains to be explored.
In accordance with previous meta-analyses [7,51,52], the incidence of NICU admission and hypoglycemia were also significantly reduced in our study. The rates of NICU admission are mainly influenced by fetal physiologic compromise, including preterm birth, hypoglycemia, respiratory distress syndrome, and neonatal jaundice. In our meta-analysis, the infants born to mothers treated with insulin needed additional management for hypoglycemia, which is partly associated with an increase in NICU admission. Neonatal hypoglycemia is one of the most common metabolic disorders of the newborn and is due to hyperinsulinemia of the fetus in response to maternal hyperglycemia in utero [49]. Fetal hypoglycemia can also lead to more serious complications such as seizures and serious brain injury [50]. Notably, metformin significantly lowered the risk of neonatal hypoglycemia by 44% in our meta-analysis, and it may reduce the risk of neonatal brain injury. The use of metformin may not harm the fetus during pregnancy and may be safer in the neonatal period with potentially beneficial effects.
A major strength of our meta-analysis is our provision of a complete overview of the effect of maternal metformin exposure on neonatal growth outcomes and neonatal adverse outcomes. We included 24 studies, which is a higher number than included in previous analyses; additionally, all of these studies were RCTs, which greatly reduced the likelihood of recall and selection biases. Moreover, a subgroup analysis by the different daily doses of metformin for treatment of GDM and an investigation of the relationship between the maternal oral dose of metformin and neonatal birth weight were carried out for the first time. Furthermore, we assessed potential publication bias by contour-enhanced funnel plots and Egger's test, the results of which suggested that our results regarding neonatal outcomes were not affected by publication bias. This increases the confidence in our findings.
Our study has several limitations that merit further discussion. First, the possibility of confounding factors in several studies cannot be completely ruled out. For example, women who had poor glycemic control with metformin and required extra insulin therapy were included in the metformin-treated group in some studies, which might cause selection bias. However, the proportion of metformin-treated women requiring insulin supplementation ranged from 8.6% to 46.8% (average, 16.2%) of the total metformin-treated women. Moreover, these patients used a lower total insulin dose than those treated with insulin alone. Therefore, we believe that such selection bias may not have influenced the overall outcomes of the studies. Second, data on neonatal growth outcomes and neonatal adverse outcomes were unavailable or incompletely reported in most of the included studies, restricting us from performing a more detailed relevant analysis and obtaining more comprehensive results. Finally, although subgroup and sensitivity analyses were performed to explore the potential sources of heterogeneity in neonatal birth weight, the cause of the high heterogeneity remains unclear.
In conclusion, the results of this meta-analysis add to the evidence that metformin may be particularly useful in women with GDM at high risk for neonatal hypoglycemia, women who want to limit maternal and fetal weight gain, or women with an inability to afford or use insulin safely. Metformin can effectively lower neonatal birth weight and the incidences of macrosomia, neonatal hypoglycemia, and NICU admission compared with insulin without an increased risk of neonatal adverse outcomes. Whether the effect of metformin on neonatal birth weight is associated with the oral dose of metformin requires further investigation in large-scale trials.

Ethical Standard Statement
We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us.
All procedures in this study were conducted in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975. We confirm that we have given due consideration to the protection of intellectual property associated with this work and that there are no impediments to publication, including the timing of publication, with respect to intellectual property. In so doing we confirm that we have followed the regulations of our institutions concerning intellectual property.
As the corresponding author, I confirm on behalf of all authors that there have been no involvements that might raise the question of bias in the work reported or in the conclusions, implications, or opinions stated.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.