Recibido: 28 de mayo de 2024; Aceptado: 17 de octubre de 2025
Metformin vs. insulin in gestational diabetes mellitus: A systematic review and meta-analysis of randomized studies
Metformina frente a insulina en la diabetes mellitus gestacional: una revisión sistemática y metaanálisis de estudios aleatorizados
Abstract:
Context:
Gestational diabetes mellitus (GDM) is defined as any degree of glucose intolerance with onset or first recognition during pregnancy, with most cases resolving after delivery. Adequate glycemic control is essential to reduce the risk of maternal and fetal complications. Historically, insulin has been the standard therapy, but recently, metformin has emerged as an alternative treatment.
Objectives:
This meta-analysis aims to evaluate maternal outcomes in gestational diabetes mellitus patients treated with insulin versus metformin.
Methods:
A systematic search was performed in PubMed, Embase, and Cochrane databases. Randomized controlled trials (RCTs) were included.
Results:
Nineteen randomized controlled trials comprising 4,320 patients were analyzed, evaluating six maternal outcomes. Metformin was associated with a reduced risk of gestational hypertension (RR: 0.65; 95% CI: 0.49-0.87; P=0.77; I²=0%) and preeclampsia (RR: 0.57; 95% CI: 0.46-0.72; P=0.08; I²=39%). The risk of lower cesarean section rates (LCLS) was also reduced with metformin (RR: 0.92; 95% CI: 0.85-0.99; P=0.0002; I²=63%). Induced labor was less frequent with metformin (RR: 0.85; 95% CI: 0.76-0.95; P=0.01; I²=57%). The incidence of spontaneous vaginal delivery was higher in the metformin group (56.1%) (RR: 1.09; 95% CI: 1.03-1.17; P=0.002; I²=61%). No statistically significant difference was found in preterm birth rates between groups (RR: 0.91; 95% CI: 0.74-1.13; P=0.002; I²=60%).
Conclusion:
This meta-analysis provides evidence supporting the use of metformin in the management of gestational diabetes mellitus, showing significant benefits in reducing rates of preeclampsia, gestational hypertension, and cesarean delivery, while also increasing spontaneous vaginal deliveries. Insulin may still be required in selected cases but appears to be associated with less favorable maternal outcomes.
Keywords:
Metformin, Insulin, Glucose intolerance, Pregnancy, Preeclampsia, Risk, Incidence..Resumen
Contexto:
la diabetes mellitus gestacional (DMG) se define como cualquier grado de intolerancia a la glucosa con inicio o primer reconocimiento durante el embarazo, la mayoría de los casos se resuelven al finalizar la gestación. El control adecuado de la glucosa es esencial para reducir el riesgo de complicaciones. Tradicionalmente, se ha utilizado la insulina como tratamiento estándar, sin embargo, recientemente la metformina ha emergido como una alternativa terapéutica.
Objetivos:
este estudio busca evaluar los resultados maternos en pacientes con diabetes mellitus gestacional tratados con insulina o metformina.
Metodología:
se realizó una búsqueda sistemática en PubMed, Embase y Cochrane, incluyendo ensayos clínicos aleatorizados (ECAs).
Resultados:
la presente revisión incluyó 19 ensayos clínicos aleatorizados con un total de 4320 pacientes y analizó seis desenlaces maternos. La hipertensión inducida por el embarazo fue menor en el grupo tratado con metformina (RR: 0,65; IC 95 %: 0,49-0,87; = 0,77; I² = 0 %), al igual que la preeclampsia (RR: 0,57; IC 95 %: 0,46-0,72; p = 0,08; I² = 39 %) y la incidencia de cesárea (LCLS) (RR: 0,92; IC 95 %: 0,85-0,99; p = 0,0002; I² = 63%). El parto inducido también fue menor con metformina (RR: 0,85; IC 95 %: 0,76-0,95; p = 0,01; I² = 57 %). La incidencia de parto vaginal normal fue mayor en el grupo de metformina (56,1 %) (RR: 1,09; IC 95 %: 1,03-1,17; p = 0,002; I² = 61 %). No se encontraron diferencias estadísticas significativas entre grupos en el parto prematuro (RR: 0,91; IC 95 %: 0,74- 1,13; p = 0,002; I² = 60 %).
Conclusiones:
Este metaanálisis presenta evidencia del uso de metformina en el tratamiento de la diabetes mellitus gestacional, mostrando diferencias significativas en la reducción de las tasas de preeclampsia, hipertensión gestacional y partos por cesárea, al tiempo que aumenta los partos vaginales espontáneos. La insulina puede utilizarse de manera complementaria si es necesario, aunque parece estar asociada con resultados maternos menos favorables.
Palabras clave:
metformina, insulina, intolerancia a la glucosa, embarazo, preeclampsia, riesgo, incidencia..Introduction
Gestational diabetes mellitus (GDM) is defined as any degree of glucose intolerance with onset or first recognition during pregnancy (1), which, in most cases, resolves at the end of gestation (2). Gestational diabetes affects different racial groups (3), with a higher prevalence reported in South Asian countries (4). Women with GDM must achieve strict glycemic control to avoid the complications associated with hyperglycemia (5).
In past years, insulin was the primary method used to control glycemia during pregnancy, reduce the risks associated with hyperglycemia, and maintain normal glucose levels. However, an alternative medication was deemed necessary- one that would be safe and effective for both mother and infant, as well as acceptable and more affordable for women. Metformin, a biguanide, emerged as such an option, as it reduces insulin resistance and lowers glucose levels. Metformin acts to reduce insulin resistance and improves insulin sensitivity through activation of AMP kinase and decreasing ATP concentrations in hepatocytes (6). Insulin sensitivity and hyperglycemia are improved by reducing hepatic gluconeogenesis and increasing peripheral glucose uptake and utilization. Metformin also decreases markers of endothelial activation, which are closely associated with insulin resistance (7,8). This biguanide crosses the placenta and acts as an insulin sensitizer and, therefore, does not cause neonatal hypoglycemia (9). Moreover, metformin has shown no evidence of teratogenicity in animal models or in human studies, including pregnancies complicated by polycystic ovary syndrome (10-16).
Concerns regarding maternal outcomes associated with insulin or metformin use in gestational diabetes mellitus have been addressed in clinical trials and meta-analyses. However, in recent years, a considerable number of new clinical trials have been published. Therefore, this meta- analysis seeks to evaluate recent publications and to compare insulin and metformin in the treatment of gestational diabetes mellitus, with a focus on maternal outcomes, to determine which of these medications provides better results.
Methods
Study selection
The systematic review and meta-analysis were conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and the Cochrane Collaboration recommendations (17). A bibliographic search was performed in PubMed, Cochrane, and Embase. The search strategy applied was: (“metformin” OR “biguanides” OR “hypoglycemic”) AND (“insulin” OR “insulin therapy” OR “insulin analogues”) AND (“gestational diabetes” OR “GDM”) AND (“RANDOM” OR“RCT”). The review protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO), with registration number CRD42023480992.
Eligibility criteria
The inclusion criteria were based on the PICO framework: P (pregnant women over 18 years old diagnosed with GDM); I (metformin); C (insulin); O (gestational hypertension, preeclampsia, LCLS, induced labor, normal vaginal delivery, preterm delivery).
The eligibility criteria included: (a) randomized trials; (b) comparison of metformin with insulin; (c) patients with GDM older than 18 years; (d) reporting at least one maternal outcome of interest (as described above), related to pregnancy and childbirth.
Exclusion criteria were: (a) studies not reporting outcomes of interest; (b) studies reporting only offspring or neonatal outcomes; (c) studies including adolescent populations; and (d) observational studies.
Study triage and data extraction
Two researchers (MFQT and AO) independently screened articles retrieved from PubMed, Embase, and Cochrane (as described in Figure 1). Selection was performed according to the predefined search strategy.
Data extraction followed the inclusion criteria, based on information from full texts and supplementary materials of eligible studies. Each investigator independently verified the accuracy of the other’s data extraction. Discrepancies were resolved through consensus. The authors collected baseline study characteristics and outcomes of interest.
Statistical analysis
Risk ratios (RR) with 95% confidence intervals (CI) were calculated for each binary endpoint. Statistical significance was defined as P<0.05. Only dichotomous outcomes were analyzed. Heterogeneity was assessed using the Cochran Q test and I² statistic; P<0.01 and I² > 25% were considered indicative of significant heterogeneity. A fixed-effects model was applied to outcomes with substantial heterogeneity. Statistical analyses were performed using Review Manager 5.4 (Cochrane Center, The Cochrane Collaboration). Forest plots were generated by the authors (MFQT).
Quality assessment - Risk of bias
The Cochrane Collaboration’s RoB-2 tool was applied to assess the risk of bias in randomized trials. Risk of bias was evaluated across five domains: bias from the randomization process, deviations from intended interventions, missing outcome data, measurement of outcomes, and selection of reported results. Each trial was classified as “low risk,” “some concerns,” or “high risk.” Risk of bias was independently assessed by four investigators (MFQT, NCJ, FSLC, and AO). A summary figure (Figure 1) was prepared by the authors to illustrate the assessment.
Results
The search strategy identified 647 records. After removal of duplicates and studies unrelated to the topic of interest based on title and abstract screening, 37 studies were fully reviewed for eligibility. Of these, 12 studies were included in the latest meta-analysis, which originally comprised 13 studies related to maternal outcomes, although one lacked available data. Ultimately, 19 randomized controlled trials (RCTs) were incorporated into the present review (Figure 1).
A total of 2,207 women were assigned to metformin treatment (with or without insulin supplementation), while 2,113 women received insulin (Table 1). The included RCTs reported different dosing regimens for both metformin and insulin (Table 2, developed by the authors MFQT and AO).
Figure 1: PRISMA flow diagram of study screening and selection
Source: Own elaboration.
Table 1: Studies included in this meta-analysis and corresponding patients assigned to each treatment
Author, year
Country
Participants
Metformin
Insulin
Ainuddin, 2015 (16)
Pakistan
Women aged 20-46 years; 20-36 weeks
75
75
Wasim, 2019 (18)
Pakistan
22-34 weeks - singleton pregnancy
137
141
Khan, 2017 (19)
Pakistan
Women aged 24, 92 - 30, 54 years; 27,32- 32,19 weeks
385
385
Huhtala, 2020 (20)
Finland
Women aged 31,8 - 37,2 years
98
97
Saleh, 2016 (21)
Egypt
24-34 weeks - singleton pregnancy
67
70
Hassan, 2012 (22)
Pakistan
Women aged 18-45 years; 20-36 weeks - singleton pregnancy
75
75
Ashoush, 2016 (23)
Egypt
Singleton pregnancy
48
47
Eid, 2018 (24)
Egypt
Women aged 18-42 years; 22-30 weeks - singleton pregnancy
113
116
Niromanesh, 2012 (25)
Iran
Women aged 18-45 years; 20-34 weeks - singleton pregnancy
80
80
Ijas, 2011 (26)
Finland
Women aged 18-45 years; 12-34 weeks, singleton pregnancy
47
50
Rowan, 2008 (27)
New Zeland and Australia
Woman aged 18-45 years; 20-33 weeks - singleton pregancy
373
378
Muhsen, 2022 (28)
Iraq
Woman, 12-34 weeks - singleton pregancy
50
50
Galal, 2019 (29)
Egypt
Woman, 28-34 weeks - singleton pregancy
56
50
Tertti, 2013 (30)
Finland
Woman, 22-34 weeks - singleton pregancy
110
107
Rodrigues, 2020 (31)
Portugal
Woman, 27-34 weeks
94
41
Picón-César, 2021 (32)
Spain
Woman aged 18-45 years; 14-35 weeks - singleton pregnancy
100
100
Spaulonci, 2013 (33)
Brazil
Woman, 24-34 weeks, singleton pregnancy
47
47
Ghomian, 2018 (34)
Iran
Woman aged 18-40 years; 24-28 weeks - singleton pregnancy
143
143
Rönnemaa, 2014 (35)
Finland
Woman, 24-32 weeks
110
107
Source: Own elaboration.
Table 2: Doses used in each RCT in metformin and insulin
Author, year
Metformin (mg)
Insulin(units)
Ainuddin, 2015 (16)
500-2500
Not Reported
Wasin, 2019 (18)
500-2500
0,7-0,8 U Kg
Khan, 2017 (19)
500- 1000
0,7 U Kg
Huhtala, 2020 (20)
500-2000
Not Reported
Saleh, 2016 (21)
500-3000
Not Reported
Hassan, 2012 (22)
500-3000
Not Reported
Ashoush, 2016 (23)
1000-2500
Not Reported
Eid, 2018 (24)
500-2500
Not Reported
Niromanesh, 2012 (25)
1000-2500
Not Reported
Ijas, 2011 (26)
750-2250
30
Rowan, 2008 (27)
500-2500
42
Muhsen, 2022 (28)
850
Not Reported
Galal, 2019 (29)
500-2000
0,8 U Kg
Tertti, 2013 (30)
500-1000
Not Reported
Rodrigues, 2020 (31)
Not Reported
Not Reported
Picón-César, 2021 (32)
425-2550
0,2 U Kg (determir) 0,1 U kg (aspart)
Spaulonci, 2013 (33)
1700-2550
0,4 U Kg
Ghomian, 2018 (34)
500-1500
0,1 U Kg
Rönnemaa, 2014 (35)
500-2000
Not Reported
This review analyzed six maternal outcomes in relation to gestational diabetes mellitus: gestational hypertension, caesarean (LCLS), preeclampsia, preterm delivery, induced labor, and normal vaginal delivery.
The forest plot for gestational hypertension favored metformin (RR: 0.65; 95% CI: 0.49-0.87; P=0.77; I²=0%), with a higher incidence observed among insulin-treated patients. Heterogeneity analysis showed I²=0%, P>0.01, and a low Chi² value (5.64). The overall effect was Z=2.93, P=0.003 (P<0.05, as defined in the methods section), indicating a small effect size for this subgroup. Hypertension occurred in 73 of 1,594 patients (4.57%) receiving metformin (Figure 2), compared with 111 of 1,539 patients (7.21%) treated with insulin alone.
The forest plot for preeclampsia also favored metformin (RR: 0.57; 95% CI: 0.46-0.72; P=0.08; I²=39%). Heterogeneity was moderate, with I²=39%, P<0.01, and Chi²=18.15. The overall effect was Z=4.75, P<0.00001, suggesting a strong effect despite moderate heterogeneity. Preeclampsia was reported in 104 of 1,790 patients (5.8%) treated with metformin, compared with 182 of 1,747 patients (10.4%) receiving insulin (Figure 3).
The forest plot for caesarean favored metformin (RR: 0.92; 95% CI: 0.85-0.99; P=0.0002; I²=63%), with more cases occurring in the insulin group. Heterogeneity was moderate (I²=63%, P<0.01, Chi²=45.75). The overall effect was Z=2.14, P=0.03, indicating a small effect size. LCLS occurred in 678 of 1,834 patients (36.96%) treated with metformin, compared with 722 of 1,784 patients (40.47%) treated with insulin (Figure 4).
The forest plot for induced labor favored metformin (RR: 0.85; 95% CI: 0.76-0.95; P=0.01; I²=57%). Heterogeneity was moderate (I²=57%, P=0.01, Chi²=20.77). The overall effect was Z=2.79, P=0.005, indicating a strong effect with moderate heterogeneity. Induced labor occurred in 340 out of 866 patients (39.2%) in the metformin group, compared with 370 out of 820 patients (45.1%) treated with insulin (Figure 5).
The forest plot for normal vaginal delivery (NVD) favored metformin (RR: 1.09; 95% CI: 1.03- 1.17; P=0.002; I²=61%), with a higher incidence among patients receiving metformin. Heterogeneity was moderate (I²=61%, P<0.01, Chi²=30.52). The overall effect was Z=2.72, P=0.007, suggesting a strong effect despite heterogeneity. NVD occurred in 842 of 1,500 patients (56.1%) treated with metformin, compared with 740 of 1,452 patients (51.0%) treated with insulin (Figure 6).
The forest plot for preterm delivery showed no significant difference between groups (RR: 0.91; 95% CI: 0.74-1.13; P=0.002; I²=60%).
Heterogeneity was moderate (I²=60%, P<0.01, Chi²=32.73). The overall effect was Z=0.86, P=0.39, indicating no meaningful effect. Preterm delivery occurred in 151 of 1,799 patients (8.39%) receiving metformin and in 162 of 1,758 patients (9.21%) treated with insulin (Figure 7).
Figure 2: Induced hypertension
Figure 3: Preeclampsia
Figure 4: LCLS
Figure 5: Induced labor
Figure 6: Normal vaginal delivery
Figure 7: Preterm delivery
Risk of bias
In this review, no study was classified as having “low risk” of bias. The majority of the studies (n=17) were categorized as “high risk,” namely: Wasim, 2019 (18); Khan, 2017 (19); Saleh, 2016(21); Hassan, 2012 (22); Ashoush, 2016 (23); Eid, 2018 (24); Niromanesh, 2012 (25); Ijas, 2010 (26); Rowan, 2008 (27); Muhsen, 2022 (28); Galal, 2019(29); Picón-César, 2021 (32); Rodrigues, 2020(31); Spaulonci, 2013 (33); Rõnnemaa, 2014 (35), and Tertti, 2013 (30).
Most of the studiesresulted in “high risk”, namely: Wasim, 2019 (18); Khan, 2017 (19); Saleh, 2016 (21); Hassan, 2012 (22); Ashoush, 2016 (23); Eid, 2018 (24); Niromanesh, 2012(25); Ijas, 2010 (26); Rowan,2008 (27); Muhsen, 2022 (28); Galal, 2019 (29); Picón-César, 2021(32); Rodrigues, 2020 (31); Spaulonci, 2013 (33); Rõnnemaa, 2014 (35), and Tertti, 2013 (30).
Finally, three studies were classified as presenting“some concerns”: Ainuddin, 2015 (16); Huhtala, 2021 (20), and Ghomian, 2018 (34) (Table 3).
Source: Own elaboration.
Table 3: Risk of bias
Study
Bias from randomization process
Bias due to deviations from intended interventions
Bias due to missing outcome data
Bias in measurement of the outcomes
Bias in selection of reported result
Overall risk of bias
Ainuddin, 2015 (16)
Low
Some Concerns
Low
Low
Some Concerns
Some Concerns
Wasin, 2019 (18)
Low
Some Concerns
Low
Some Concerns
High
High
Khan, 2017 (19)
Low
Low
Low
Some Concerns
High
HIgh
Huhtala, 2021 (20)
Low
Low
Low
Some Concerns
Some Concerns
Some Concerns
Saleh, 2016 (21)
Low
Low
Low
Some Concerns
High
High
Hassan, 2012 (22)
Low
Low
Low
Some Concerns
High
High
Ashoush, 2016 (23)
Low
Low
Low
Some Concerns
High
High
Eid, 2018 (24)
Low
Low
Low
Some Concerns
High
High
Niromanesh, 2012 (25)
Low
Low
Low
Some Concerns
High
High
Ijas, 2010 (26)
Low
Some Concerns
Low
High
High
High
Rowan, 2008 (27)
Low
Low
Low
High
High
High
Muhsen, 2022 (28)
Low
High
Low
High
High
High
Galal, 2019 (29)
Low
Some Concerns
Low
High
High
High
Picón-César, 2021 (32)
High
Low
Some concerns
High
High
High
Rodrigues, 2020 (31)
High
Low
Low
High
High
High
Ghomian, 2018 (34)
Some concerns
Low
Low
Low
Some concerns
Some concerns
Spaulonci, 2013 (33)
Low
Low
Low
Some Concerns
High
High
Rönnemaa, 2014 (35)
Low
High
Low
High
Some concerns
High
Tertti, 2013 (30)
Low
Some Concerns
Low
High
High
High
Discussion
In this meta-analysis, a total of 4,320 patients were included across 19 randomized controlled trials. Six outcomes were statistically compared:
(1) induced hypertension occurred more frequently in patients receiving insulin; (2) preeclampsia was more prevalent in the insulin group, with 10.4% versus 5.8% in the metformin group; (3) large-for- gestational-age infants (LGA) were more common in the insulin group; (4) induced labor occurred significantly more in the insulin group; (5) normal vaginal delivery was more frequent in patients treated with metformin; and (6) preterm delivery was more common in the insulin group.
A previous meta-analysis by Bao et al. (36) evaluated 1,457 patients, focusing on induced hypertension, LCLS, gestational age at delivery, maternal weight gain, and premature delivery. In comparison, this meta-analysis included a larger cohort (4,367 patients) and focused specifically on maternal outcomes and modes of labor. Both meta-analyses consistently demonstrated that metformin was associated with lower rates of induced hypertension (3).
Regarding preeclampsia, the previous meta- analysis did not identify a statistically significant difference; however, the present study observed a trend favoring metformin, reinforcing its potential protective effect against hypertensive disorders in pregnancy (36, 24, 26, 27). LCLS also demonstrated a statistically significant reduction in the metformin group (P=0.0002) (20, 22, 24, 27, 32).
Importantly, the safety profile of metformin during pregnancy is well established. Several studies indicate that metformin crosses the placenta but does not induce teratogenic effects in humans or animal models (6-12). Long-term follow-up studies have evidenced no adverse outcomes in offspring, including growth, metabolic parameters, or neurodevelopment, up to early childhood (20, 22, 27). These findings support the use of metformin as a safe alternative to insulin, particularly in patients with gestational diabetes mellitus requiring glycemic control.
Neonatal implications of metformin use have been thoroughly evaluated. Evidence suggests that metformin does not increase the risk of neonatal hypoglycemia, respiratory distress, or admission to neonatal intensive care units and may reduce the incidence of macrosomia (20, 22, 27). Furthermore, metformin exposure has not been associated with adverse long-term metabolic outcomes, and children exposed in utero exhibit growth trajectories and body composition comparable to those of insulin-exposed or unexposed infants (20, 22, 27).
In three RCTs included in this review, metformin combined with insulin did not differ significantly from metformin alone, reinforcing its versatility in clinical practice (17, 19, 23). Therefore, metformin can be recommended as a first-line pharmacological therapy for GDM, with the addition of insulin reserved for cases where glycemic targets are not achieved (17-37).
Overall, this meta-analysis strengthens current evidence that metformin is not only effective in controlling maternal glycemia but also safe for both the mother and neonate in the short and long term, supporting its broader implementation in the management of GDM (6-12,17-37).
Conclusion
This systematic review and meta-analysis of nineteen randomized controlled trials evaluated maternal outcomes and modes of labor in gestational diabetes mellitus treated with metformin and insulin. The analysis demonstrates that insulin therapy is significantly associated with higher rates of large-for-gestational-age infants, induced labor, and preeclampsia, whereas metformin treatment is linked to a higher incidence of normal vaginal deliveries. These findings support the use of metformin as an effective and safe alternative to insulin for glycemic management in GDM, offering favorable maternal outcomes. Nonetheless, combination therapy with insulin should remain an option for patients who do not achieve adequate glycemic control with metformin alone, allowing for individualized treatment and the optimization of both maternal and neonatal outcomes.
Authors’ contributions
Maria Fernanda Quandt Treml: Conceptualization, formal analysis, research, writing (original draft); Nicole Caroline Junglos: Formal analysis, research, methodology, writing (original draft); Agelicia Ott: Formal analysis, writing (original draft); Heloisa Bernardi Hummel: Formal analysis, writing (original draft); Ana Carolina Moreira de Moraes Lima: Formal analysis, writing (original draft); Felipe Silva Luciano Carvalho: Formal analysis, writing (original draft); Matheus José Barbosa Moreira: Writing (referee and editing corrections).
Ethical implications
This article had no ethical implications.
Funding
This article had no financial support.
Conflicts of interest
None of the authors have any conflict of interest to disclose.
AI disclosure statement
The authors declare that no artificial intelligence tools were used in the preparation or writing of this manuscript.
Availability of data and materials
No data are available in a public repository. For inquiries regarding any information related to this article, please contact the corresponding author.