Optimizing diabetes care in pregnancy: A systematic review of glycemic targets, treatment initiation, and glucose‐lowering pharmacological management for T1DM, T2DM, and GDM

Abstract

Introduction

Pregnant women with diabetes are at increased risk of adverse maternal and neonatal outcomes, yet the optimal approach to achieving recommended glucose levels and glucose‐lowering pharmacological strategies during pregnancy remains unclear.

Methods

We conducted a systematic review with meta‐analysis of randomized controlled trials (RCTs) synthesizing the evidence on glycemic targets and pharmacological interventions for pregnant women with pre‐existing diabetes or gestational diabetes mellitus (GDM). We extracted data for outcomes prioritized for the development of all WHO guidelines on maternal and perinatal health, as well as intervention‐specific outcomes. The selected critical outcomes were maternal death, maternal functioning and women's views and experiences, stillbirth/fetal death, neonatal death, and perinatal death. We followed standard Cochrane methods.

Results

The evidence on glycemic targets was very limited. Among women with type 1 diabetes mellitus (T1DM), meeting stricter glycemic targets may increase the risk of hypoglycemia and longer hospitalization, whereas less stringent targets may increase the risk of caesarean birth, large‐for‐gestational‐age babies, and neonatal respiratory distress syndrome. In women with GDM, no clear differences were observed between a tight and moderate target. Evidence on the optimal approach to glucose‐lowering pharmacotherapy for the management of type 1 and type 2 diabetes during pregnancy was also limited. In women with GDM, metformin was associated with better maternal and neonatal outcomes than glibenclamide. Compared to insulin, metformin probably reduces the risk of neonatal intensive care admission and neonatal hypoglycemia. We rated most outcomes as low or very low certainty, which means that the true effects of the interventions may differ substantially from current estimates, and new evidence is likely to change our confidence in the estimates of effects. The main reasons for downgrading the certainty of the evidence were risk of bias and imprecision.

Conclusions

This review did not yield sufficient evidence to determine which glycemic targets are the most effective in improving health outcomes in women with pre‐existing diabetes or GDM. Current evidence on glucose‐lowering pharmacotherapy in pregnancy is limited and generally of low certainty. In women with GDM, metformin and insulin appeared to offer benefits compared to glibenclamide, and metformin may be a safe first‐line alternative to insulin, although the evidence is limited for most outcomes. High‐quality RCTs are needed to clarify the effects of these interventions, including into the longer term for the child.

PROSPERO registration

CRD42025630036

1. INTRODUCTION

1.1. Description of the condition

In a recent meta‐analysis [1], the global pooled prevalence of gestational diabetes mellitus (GDM) was 14.0%. Differences in the prevalence of GDM were observed across income levels, with standardized prevalences of 12.7%, 9.2%, and 14.2% in low‐, middle‐, and high‐income settings, respectively.

Another meta‐analysis [2] examined the prevalence of pre‐existing diabetes in pregnancy. The pooled estimate for type 2 diabetes was 0.2%, ranging from 0.1% in Sweden to 2.9% in Saudi Arabia. Regional differences were also observed, with the lowest in North America (0.2%) and the highest in the Middle East and North Africa (1.5%). The pooled prevalence of type 1 diabetes was 0.3%, with values between 0.1% and 0.5%. Prevalence was similar in Australasia, North America, and Europe (0.3%) but higher in the Middle East and North Africa (0.9%).

Women with diabetes are at an increased risk of complications during pregnancy and at delivery. In addition, women with GDM and possibly their children are also at increased risk of type 2 diabetes in the future [3].

1.2. Description of the interventions

There is no globally accepted standard from the World Health Organization (WHO) for pregnancy glycemic targets, and countries often adopt international guidance or their own consensus values. The American Diabetes Association (ADA) [4] and the National Institute for Health and Care Excellence (NICE) [5] recommend fasting plasma glucose <5.3 mmol/L (<95 mg/dL) and 1‐h postprandial glucose <7.8 mmol/L (<140 mg/dL); however, NICE's 2‐h postprandial glucose is stricter (ADA: <6.7 mmol/L [<120 mg/dL] vs. NICE: <6.4 mmol/L [<115 mg/dL]). The Australasian Diabetes in Pregnancy Society (ADIPS) [6] recommends fasting 4.0–5.5 mmol/L (72–99 mg/dL); 1‐h post‐prandial <8.0 mmol/L (<144 mg/dL); and 2‐h post‐prandial <7 mmol/L (<126 mg/dL).

Regarding glucose‐lowering pharmacological management, there is also variability in the recommendations. In women with type 1 diabetes, there is an established consensus that insulin should be continued and adjusted as needed [4, 5, 7]. There is also consensus on the use of multiple daily injections or pump therapy (this may be the only option in many contexts), but NICE [5] specifically recommends offering pump therapy to women who are already on multiple daily injections and not meeting glycemic targets. There is less clarity regarding the preferred insulin regimen.

In women with type 2 diabetes who are already on oral glucose‐lowering medication, and women with GDM who cannot achieve adequate glycemic goals with lifestyle changes alone, there is no international consensus on which medication should be used as first‐line. The American College of Obstetricians and Gynecologists [7] recommends insulin as first‐line treatment, but allows for the use of metformin and glibenclamide as alternatives; whereas ADA [4] recommends against using metformin and glibenclamide, individually or in combination, as first‐line agents for management of diabetes in pregnancy, as both cross the placenta to the fetus and may not be sufficient to achieve glycemic targets. In contrast, NICE [5] recommends metformin as first‐line treatment, with insulin used instead or added if metformin is contraindicated, unacceptable to the woman, or if the woman fails to meet glucose targets. These inconsistencies reflect the need for an up‐to‐date review to inform the development of new WHO recommendations that can be adapted to all settings globally.

1.3. Why it is important to do this review

With the increasing prevalence of type 1 diabetes mellitus (T1DM), type 2 diabetes mellitus (T2DM), and GDM, there is an urgent need for evidence‐based strategies to support meeting glycemic targets during pregnancy and improve maternal and neonatal outcomes. It is also necessary to identify the most appropriate first‐line pharmacological treatment in pregnant women with T2DM or GDM, and the most suitable insulin regimen for those with T1DM. It is therefore essential to critically assess the current evidence regarding the benefits and potential harms of various glucose‐lowering agents in pregnant women with diabetes.

2. OBJECTIVE

The objectives of this review are (1) to evaluate the effects of different glucose targets in pregnant women and (2) to compare different glucose‐lowering pharmacological approaches for improving maternal and neonatal outcomes.

3. MATERIALS AND METHODS

This systematic review and meta‐analysis follows standard Cochrane methodology [8] and adheres to Preferred Reporting Items for Systematic reviews and Meta‐Analyses (PRISMA) guidelines [9]. We used DistillerSR software (DistillerSR v2.35) for screening and data extraction, RevMan Web for data analysis (RevMan v9.7.1), and GRADEpro for creating Evidence Profiles.

3.1. Criteria for considering studies for this review

3.1.1. Types of participants

We included studies of pregnant women with diabetes, including women with pre‐existing type 1 or type 2 diabetes and women with GDM. We analyzed population subgroups separately according to the type of diabetes.

We excluded studies that included mixed populations if results were not disaggregated.

3.1.2. Types of interventions

• For pregnancy‐specific glycemic targets, we included any glycemic targets, as measured by self‐monitoring of blood glucose (SMBG) or continuous glucose monitoring (CGM).

• For glycemic targets as triggers for medication, we included any specific glycemic management target used as a trigger for initiating or escalating medication (new oral glucose‐lowering agent or new type of insulin).

• For pharmacological interventions for the management of type 1 diabetes, we included studies comparing (1) different types of insulin regimens; (2) different insulin delivery systems; (3) the adjunctive use of metformin with insulin.

• For pharmacological interventions for the management of type 2 diabetes and GDM, we included any oral or insulin regimen, alone or in combination, compared to (1) placebo, (2) diet and/ or physical activity, or (3) any other treatment (i.e., head‐to‐head comparisons of glucose‐lowering pharmacological interventions).

3.1.3. Types of outcomes

A standard set of outcomes was predefined and consistently applied across all systematic reviews supporting the development of the WHO guideline (see Table 1). In addition, each review included outcomes specific to the intervention (see Table 2).

TABLE 1.

Standard maternal, neonatal, and health system outcomes—all reviews.

Newborn	Maternal	Health system
Stillbirth/fetal death Neonatal death Perinatal death Admission to neonatal intensive care Apgar score of <7 at 5 min Birthweight Gestational age at birth Intrauterine growth restriction Low birthweight (<2500 g) Preterm birth (<32 weeks; <37 weeks) Small‐for‐gestational age (<10th centile for gestational age) Neurodevelopmental outcomes (into infancy and childhood)	Maternal death Mode of birth Women's views and experiences Admission to intensive care Antenatal hospitalization Maternal functioning/well‐being	Antenatal admissions to the hospital for the woman and the length of stay Emergency department visits for the woman Length of postnatal hospitalization for the woman Length of neonatal hospitalization Costs of care for the woman or baby or both

TABLE 2.

Additional outcomes—pharmacological management of T1DM, T2DM, and GDM.

Newborn	Maternal
Birth injury Congenital anomalies Hyperbilirubinemia Large‐for‐gestational age Macrosomia Neonatal hypoglycemia Neonatal seizures Respiratory distress syndrome Shoulder dystocia Transient tachypnoea of the newborn Long‐term cardiometabolic outcomes	Childbirth‐related injuries Early pregnancy loss Gestational weight change Hypertensive disorders of pregnancy Maternal glycemia Need for additional (glucose‐lowering) agents or medication change Induction of labor Long‐term cardiometabolic outcomes

Abbreviations: GDM, gestational diabetes mellitus; T1DM, type 1 diabetes mellitus; T2DM, type 2 diabetes mellitus.

WHO selected the following critical outcomes:

• Maternal outcomes: maternal death, maternal functioning/well‐being, and women's views and experiences.

• Neonatal outcomes: stillbirth/fetal death, neonatal death, and perinatal death.

All other outcomes were rated as important.

3.1.4. Types of studies

Randomized controlled trials (RCTs) were prioritized for inclusion.

3.2. Search methods for identification of studies

An electronic search was conducted in the following databases: the Cochrane Central Register of Controlled Trials (CENTRAL), Medline (OVID), Embase (OVID), and LILACS. The searches were conducted in November 2024. No restrictions were applied regarding date of publication, publication status (i.e., published, unpublished, in press, and in progress), or language.

In addition, an overarching search for systematic reviews and clinical practice guidelines was conducted in July 2024. The reference list of relevant reviews was screened for potentially eligible studies.

3.3. Selection of studies

Two review authors independently screened all citations and abstracts identified by the search. We obtained full reports for potentially eligible studies, and these were also independently screened by two review authors. We resolved any disagreements by consensus or by involving a third reviewer or content expert.

We restricted the inclusion criteria to published, peer‐reviewed studies. We conducted a “studification” procedure to ensure that all publications related to the same study were identified and grouped together (such as protocols, conference abstracts, trial registrations, and companion reports) to avoid duplication of data.

3.4. Data extraction and management

One reviewer extracted data using pre‐tested data extraction forms. A second reviewer cross‐checked the extracted data. We resolved any disagreements about data extraction by referring to the study report and through discussion.

For each included study, we extracted data on study setting, funding, conflict of interest, inclusion and exclusion criteria, participants’ diagnosis and characteristics, intervention details, and outcome data.

3.5. Assessment of risk of bias in included studies

We assessed the risk of bias of each RCT using the Cochrane Risk of Bias 2 (RoB 2) tool [10]. This tool evaluates potential bias across five domains: (1a) bias arising from the randomization process (adequacy of random sequence generation and allocation concealment); (1b) bias arising from period and carryover effects (whether the design and analysis appropriately address potential effects from previous intervention periods, for cross‐over trials only); (2) bias due to deviations from intended intervention (whether participants and personnel adhered to the assigned intervention and if deviations were balanced), (3) bias due to missing outcome data (extent and handling of incomplete outcome data); (4) bias in measurements of the outcome (appropriateness and consistency of outcome measurement methods and blinding of outcome assessors); and (5) bias in the selection of the reported result (whether reported analyses correspond to pre‐specified plans and outcomes).

Each domain was judged as having low risk of bias, some concerns, or high risk of bias for each specific outcome.

3.6. Research integrity assessment

We applied the Research Integrity Assessment (RIA) tool to evaluate the trustworthiness of included trials [11, 12]. This tool includes the following six domains: (1) retraction or expression of concern (whether the study had been retracted or was subject to a formal expression of concern); (2) prospective trial registration (whether the trial was registered in a recognized registry before enrolment of the first participant); (3) adequate ethics approval and informed consent (whether approval from a nationally recognized ethics committee and written informed consent were obtained and reported); (4) author group (plausibility of author affiliations, number of authors, and consistency with the reported study location); (5) sufficient reporting and plausibility of methods (adequacy and clarity of reporting of the study design—e.g., randomization—and plausibility of the described methods); and (6) plausibility of results (plausibility of recruitment rates, participant characteristics, effect sizes, and data consistency).

We used a modified approach to handling studies based on the RIA assessment.

• Trials with critical concerns (e.g., retracted, subject to expressions of concerns, or with severe issues in the methodological conduct and/or plausibility of the results) were excluded from the analyses.

• Trials with no or minor concerns were included in the main analyses.

• For trials with serious but not critical concerns, we attempted to contact the study authors for clarification; depending on their response, these studies were reclassified to the above categories. In case of unsatisfactory response, the studies were excluded from the review. In case of no response, the studies were included in the main analysis and excluded from the sensitivity analysis; where exclusion led to a substantially different result (different direction of effect or level of certainty), the sensitivity analysis result was reported instead. Sensitivity analysis was not feasible when a single trial was included in the analysis.

3.7. Measurement of treatment effect

For dichotomous outcomes, results were reported as risk ratios (RRs) with their respective 95% confidence intervals (CIs). For continuous data, we used mean differences (MDs). For count/rate outcomes, we planned to use the rate ratio with 95% CI as the effect measure.

3.8. Assessment of heterogeneity

We assessed heterogeneity by visually inspecting the forest plot (non‐overlapping CIs generally signify statistical heterogeneity) and by using the I ² statistic, which describes the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error.

3.9. Data synthesis

Where possible, we meta‐analyzed the data using a random‐effects model. Where we considered meta‐analysis to be inappropriate or not feasible, we summarized data in tables. Outcomes from cross‐over trials (mostly related to glycemic control, such as time in range) were reported narratively.

3.10. Assessment of the certainty of the evidence

The results of the reviews are presented in Grading of Recommendations, Assessment, Development and Evaluations (GRADE) Evidence Profiles, rating the certainty of the evidence according to the Cochrane Handbook [8].

The certainty of the evidence started at high, and it could be downgraded for (1) risk of bias (limitations in study design and conduct of the study); (2) inconsistency (unexplained heterogeneity across the studies); (3) indirectness (difference between the research question and the included studies in terms of population, intervention, comparator, or outcomes); (4) imprecision (wide CIs or small sample sizes leading to uncertainty in the effect estimate); and (5) publication bias (suspicion that the evidence base is missing studies due to selective publication). As a result, the overall certainty of the evidence ranged between very low (we are very uncertain about the effect estimate) and high (we are confident that the true effect lies close to the estimate).

4. RESULTS

Note that this section presents only outcomes for which the certainty of the evidence was rated as low, moderate, or high. Full details on GRADE assessments are available in the GRADE Evidence Profile tables in Annexes S1–S4. The characteristics of the included studies, as well as the research integrity and risk of bias assessments, are also available in Annexes S1–S4.

4.1. Glycemic targets in pregnancy

4.1.1. Pregnant women with T1DM

Summary of included studies

We identified three RCTs evaluating different glycemic targets in pregnant women with T1DM. All studies enrolled women early in pregnancy and excluded those with significant comorbidities or advanced diabetic complications. Interventions compared varying levels of strictness in glycemic targets.

• One trial [13] (n = 137, USA, 1978–1989) compared very tight (“strict”: fasting <4.44 mmol/L [<80 mg/dL]; 1.5‐h postprandial <6.66 mmol/L [<120 mg/dL]) with tight‐moderate glycemic targets (“customary control”: fasting <5.55 mmol/L [<100 mg/dL]; postprandial <7.77 mmol/L [<140 mg/dL]).

• One trial [14] (n = 60, Saudi Arabia, date not reported) evaluated three ranges: “tight” (fasting ≤5.6 mmol/L [≤100 mg/dL]), “moderate” (fasting 5.6–6.7 mmol/L [101–121 mg/dL]), and “loose” control targets (fasting 6.7–8.9 mmol/L [121–160 mg/dL]).

• One trial [15] (n = 26, USA, 2000–2003) compared very tight (“rigid”: fasting/pre‐meal 3.33–5.00 mmol/L [60–90 mg/dL]; 1‐h postprandial 6.67–7.78 mmol/L [120–140 mg/dL]) with tight‐moderate glycemic targets (“less rigid”: fasting/pre‐meal 5.28–6.39 mmol/L [95–115 mg/dL]; postprandial 8.61–9.72 mmol/L [155–175 mg/dL]).

Risk of bias and research integrity assessment

For the three trials, at least one outcome was rated as having some concerns in the risk of bias assessment, primarily due to the lack of trial registration and insufficient information regarding the randomization process. None of the studies were identified as having significant research integrity concerns when the RIA tool was applied.

Effects of interventions

Comparison 1.1 Very tight compared to tight‐moderate glycemic targets

Maternal outcomes

Low certainty evidence from one trial suggests that compared to moderately tight glycemic targets, very tight glycemic targets may increase the percentage of days with maternal hypoglycemia (<3.3 mmol/L or 60 mg/dL) (MD, 20 days; 95% CI 7, 33; n = 22). However, there may be little to no difference in early pregnancy loss (RR, 0.17; 95% CI, 0.01, 3.23; n = 24) and maternal weight gain (MD, 0.02 kg/week; 95% CI, −0.23, 0.27; n = 22 women).

Health service outcomes

Compared to moderately tight glycemic targets, very tight glycemic targets may increase the number of days of postnatal hospitalization for the woman (MD, 10.6 more days; 95% CI, 9.53, 11.67; 1 RCT; n = 137; low certainty).

Comparison 1.2 Tight compared to moderate glycemic targets

Maternal outcomes

Tight compared to moderate glycemic targets may increase the risk of hypoglycemic episodes (RR, 26.47; 95% CI, 1.61, 435.38; 1 RCT; n = 45; low certainty).

Comparison 1.3 Moderate compared to loose glycemic targets

Maternal outcomes

Compared to loose glycemic targets, moderate glycemic targets may reduce the risk of caesarean birth (RR, 0.26; 95% CI, 0.08, 0.89; 1 RCT; n = 44 women; low certainty).

Neonatal outcomes

Compared to loose glycemic targets, moderate glycemic targets may reduce the risk of large‐for‐gestational age babies (RR, 0.02; 95% CI, 0.00, 0.31; 1 RCT; n = 44; low certainty) and respiratory distress syndrome (RR, 0.17; 95% CI, 0.04, 0.75; 1 RCT; n = 44; low certainty).

4.1.2. Pregnant women with T2DM

No trials were identified.

4.1.3. Pregnant women with GDM

Summary of included studies

We identified three RCTs that compared different glycemic targets in pregnant women with GDM.

• The TARGET trial [16] (n = 1100, New Zealand, 2015–2018, 10 clusters): “tighter” targets (fasting ≤5.0 mmol/L [≤90 mg/dL]; 1‐h postprandial ≤7.4 mmol/L [≤133 mg/dL]; 2‐h postprandial ≤6.7 mmol/L [≤121 mg/dL]) versus “less tight” targets (fasting <5.5 mmol/L [<99 mg/dL]; 1‐h postprandial <8.0 mmol/L [<144 mg/dL]; 2‐h postprandial <7.0 mmol/L [<126 mg/dL]).

• The GDM‐MOMS trial [17] (n = 60, USA, 2015–2018): “intensive” targets (fasting <5.0 mmol/L [<90 mg/dL]; 1‐h postprandial <6.67 mmol/L [<120 mg/dL]) versus standard targets (fasting <5.28 mmol/L [<95 mg/dL]; 1‐h postprandial <7.78 mmol/L [<140 mg/dL]).

• The GEM‐GDM trial [18] (n = 41, 2015, Rusia): “tight” targets (fasting <5.1 mmol/L [<92 mg/dL]; 1‐h postprandial <7.0 mmol/L [<126 mg/dL]) versus “less tight” targets (fasting <5.3 mmol/L [<95 mg/dL]; 1‐h postprandial <7.8 mmol/L [<140 mg/dL]).

Risk of bias and research integrity assessment

In all trials, at least one outcome was rated as having “some concerns” in risk of bias due to retrospective registration or unspecified outcomes, lack of blinding (for subjective outcomes only), or missing data. None of the studies were identified as having significant concerns related to research integrity as assessed using the RIA tool.

Effects of interventions

Comparison 2. Tight compared to moderate glycemic targets

Maternal outcomes

Tight versus moderate glycemic targets in women with GDM probably results in little to no effect on maternal well‐being (as assessed by the Health‐Related Quality of Life [HRQOL] 36 Item Short‐Form survey [SF‐36]) at 36 weeks (physical component—MD, 0.42; 95% CI, −1.48, 2.32; 1 RCT; n = 332; and mental component—MD, 0.20; 95% CI, −1.35, 1.75; 1 RCT; n = 332; moderate certainty); and there may be little or no difference in effect on depression at 36 weeks’ gestation (as assessed using the Edinburgh Postnatal Depression Scale [EPDS]) (RR, 1.10; 95% CI, 0.37, 3.28; 1 RCT; n = 334; low certainty); and anxiety at 36 weeks’ gestation (as assessed using the short form of the Spielberger State–Trait Anxiety Inventory) (RR, 0.85; 95% CI, 0.44, 1.63; 1 RCT; n = 327; low certainty).

High certainty evidence from one trial shows that compared to moderate glycemic targets, tight targets result in little to no difference in risk of induction of labor (RR, 0.97; 95% CI, 0.79, 1.19; n = 1097). There is also moderate certainty evidence of probably little to no difference in effect on antenatal hospitalization (RR, 1.05; 95% CI, 0.75, 1.46; 1 RCT; n = 1096), caesarean birth (RR, 1.01; 95% CI, 0.29, 3.53; 2 RCTs; n = 1138), preeclampsia (RR, 1.36; 95% CI, 0.68, 2.72; 1 RCT; n = 1096), maternal hypoglycemia (RR, 1.46; 95% CI, 0.18, 11.65; 2 RCTs; n = 1152), need for glucose‐lowering agents (RR, 1.23; 95% CI, 0.97, 1.56; 3 RCTs; n = 1197), and low certainty evidence of little to no difference on eclampsia (RR, 2.53; 95% CI, 0.10, 61.89; 1 RCT; n = 1096).

Neonatal outcomes

Tight versus moderate glycemic targets in women with GDM results in little to no difference on birthweight (MD, 11.68 g; 95% CI, 63.18, 86.54 g; 3 RCTs; n = 1202; high certainty) and on large‐for‐gestational age babies (RR, 0.98; 95% CI, 0.85, 1.13; 3 RCTs; n = 1202; high certainty). There is also probably little to no difference in effect on admission to neonatal intensive care (RR, 0.59, 0.05, 7.34; 2 RCTs; n = 1167; moderate certainty), gestational age at birth (MD, 0.11 weeks; 95% CI, −0.64, 0.86; 3 RCTs; n = 1198; moderate certainty), small‐for‐gestational age babies (RR, 0.67; 95% CI, 0.14, 3.13; 3 RCTs; n = 1202; moderate certainty), macrosomia (RR, 1.08; 95% CI, 0.68, 1.71; 1 RCT; n = 1097), shoulder dystocia (RR, 0.27; 95% CI, 0.07, 1.06; 1 RCT; n = 1097; moderate certainty), neonatal hypoglycemia (RR, 0.93; 95% CI, 0.75, 1.17; 2 RCTs; n = 1161; moderate certainty), and respiratory distress syndrome (RR, 1.16; 95% CI, 0.65, 2.07; 1 RCT; n = 1097; moderate certainty). Low certainty evidence from two trials also suggests little to no difference in neonatal hyperbilirubinemia (RR, 0.69; 95% CI, 0.17, 2.78; n = 1161).

Health service outcomes

Tight versus moderate glycemic targets in women with GDM slightly increases maternal antenatal hospital stay (MD, 0.93 days; 95% CI, 0.17, 1.69; 1 RCT; n = 1097; high certainty), but makes little to no difference in length of neonatal (MD, 0.95 days; 95% CI, 0.83, 1.07; 1 RCT; n = 1097; high certainty) and maternal postnatal hospitalization (MD, 0.98 days; 95% CI 0.88, 1.08; 1 RCT; n = 1097 women; high certainty).

4.2. Glycemic targets as triggers for medication

No trials were identified for inclusion.

4.3. Pharmacological management

4.3.1. Pregnant women with T1DM

Summary of included studies

We identified 17 trials that evaluated glucose‐lowering pharmacological treatment for type 1 diabetes during pregnancy.

Four RCTs that compared different types of insulin in pregnant women with T1DM (Comparisons 3.1–3.3).

• Two trials compared rapid‐acting insulin analogues with regular human insulin. One trial [19] (n = 33, Sweden, dates not reported) used insulin lispro, and another trial [20] (n = 322, multicenter, 2002–2005) evaluated insulin aspart. All women were also administered Neutral Protamine Hagedorn (NPH) insulin.

• One trial [21] (n = 310, multicenter, 2007–2010) compared insulin detemir with NPH insulin; both groups received insulin aspart.

• One multicenter trial [22] (n = 188, 2017–2020) compared insulin degludec versus insulin detemir; both groups received insulin aspart.

Thirteen trials compared different insulin delivery methods (Comparisons 3.4–3.7).

• Five trials [23, 24, 25, 26, 27] (n = 169, Finland, Germany, Italy, 1979–1990) compared continuous subcutaneous insulin infusion (CSII) with conventional insulin delivery. All women received human insulin.

• One cross‐over trial [28] (n = 9, UK, dates not reported) compared insulin lispro administered immediately before versus after a standard meal.

• One cross‐over trial [29] (n = 11, Belgium, 2020) compared a sensor‐augmented insulin pump in predictive low glucose suspend (PLGS) versus low glucose suspend (LGS) mode.

• Six trials compared closed‐loop insulin pump systems with conventional therapy: three parallel group RCTs [30, 31, 32] (n = 244, the Netherlands, UK, USA) and three cross‐over RCTs [33, 34, 35] (n = 46, UK). The studies evaluated different closed‐loop systems. Women in the control group received insulin by injection [30, 31], pump [30, 31, 33], or sensor‐augmented pump [32, 34, 35].

Risk of bias and research integrity assessment

Only one trial [27] was assessed as high risk of bias due to an improper randomization process. All other trials had at least one domain rated as having “some concerns.” Many trials did not adequately describe allocation concealment, and several were either unregistered, registered retrospectively, or did not specify intended outcomes. Additionally, some outcomes were rated as having some concerns due to missing outcome data or lack of blinding. No trials were identified as having significant concerns related to research integrity based on the RIA tool assessment.

Effect of interventions

Comparison 3.1 Rapid‐acting insulin analogues compared to regular human insulin

Maternal outcomes

Rapid‐acting insulin, compared to regular insulin, may increase maternal satisfaction with treatment (as assessed using the Diabetes Treatment Satisfaction Questionnaire, range: 0–100, higher scores are better) (MD, 4.2; 95% CI, 1.2, 7.2; 1 RCT; n = 322; low certainty). However, there may be little to no effect on risk of hypertensive disorders of pregnancy (RR, 0.88; 95% CI, 0.49, 1.60; 2 RCTs; n = 355; low certainty).

Neonatal outcomes

Rapid‐acting insulin, compared to regular insulin, may result in a reduction in the risk of preterm birth (defined as birth <37 weeks) (RR, 0.65; 95% CI, 0.43, 0.99; 1 RCT; n = 268 babies; low certainty), but little to no difference in gestational age at birth (MD, 0.2 weeks; 95% CI, −0.18, 0.58; 1 RCT; n = 268; low certainty).

Comparison 3.2 Long‐acting basal analogue compared to NPH

Maternal outcomes

Long‐acting basal analogue insulin, compared to NPH, may increase the risk of preeclampsia (RR, 8.32; 95% CI, 1.05, 65.70; 1 RCT; n = 310; low certainty); however, it probably results in little to no difference in risk of major hypoglycemia (RR, 0.79; 95% CI, 0.49, 1.26; 1 RCT; n = 310; moderate certainty) and possibly in little to no difference in the risk of hyperglycemia (RR, 5.20; 95% CI, 0.25, 107.36; 1 RCT; n = 310; low certainty).

Neonatal outcomes

Long‐acting basal analogue insulin, compared to NPH, may result in little to no difference in gestational age at birth (MD, 0.40 weeks; 95% CI, −0.01, 0.81; 1 RCT; n = 264 babies; low certainty) and birthweight (MD, −67 g; 95% CI, −217.63, 83.63; 1 RCT; n = 264; low certainty).

Comparison 3.3 Long‐acting basal analogues—degludec compared to determir

Maternal outcomes

Low certainty evidence from one trial suggests that compared to determir, degludec may result in little to no difference in early pregnancy loss (RR, 0.75; 95% CI, 0.25, 2.26; n = 188), planned caesarean birth (RR, 0.96; 95% CI, 0.71, 1.31; n = 185), unplanned caesarean birth (RR, 1.58; 95% CI, 0.88, 2.84; n = 185), spontaneous vaginal birth (RR, 0.60; 95% CI, 0.30, 1.19; n = 185), operative vaginal birth (RR, 0.92; 95% CI, 0.37, 2.28; n = 185), and hypertensive disorders of pregnancy (RR, 1.69; 95% CI, 0.70, 4.10; n = 171).

Neonatal outcomes

Compared to detemir, degludec may result in little to no difference in risk of preterm birth (<37 weeks) (RR, 1.51; 95% CI, 0.92, 2.47; n = 171), very preterm birth (30–34 weeks) (RR, 0.40; 95% CI, 0.08, 1.98; n = 171), large‐for‐gestational age (RR, 1.26; 95% CI, 0.97, 1.64; n = 171), small‐for‐gestational age (RR, 0.33; 95% CI, 0.03, 3.10; n = 171), neonatal hypoglycemia (RR, 1.04; 95% CI, 0.60, 1.81; n = 171), and congenital anomalies (RR, 1.04; 95% CI, 0.41, 2.66; n = 188).

Comparison 3.4 CSII compared to conventional insulin

The certainty of the evidence for all reported outcomes was judged to be very low.

Comparison 3.5 Insulin before compared to after meals

The certainty of the evidence for all reported outcomes was judged to be very low.

Comparison 3.6 Sensor‐augmented pumps: sensor‐augmented insulin pump in predictive low glucose suspend (PLGS) mode compared to sensor‐augmented insulin pump in low glucose suspend (LGS)

The certainty of the evidence for all reported outcomes was judged to be very low.

Comparison 3.7 Advanced systems: closed‐loop pump compared to conventional therapy

Maternal outcomes

Low certainty evidence suggest that, compared to conventional therapy, closed‐loop pump use may result in a reduction in the risk of caesarean birth (RR, 0.74; 95% CI, 0.59, 0.94; 3 RCTs; n = 230), gestational hypertension (RR, 0.53; 95% CI, 0.30, 0.93; 2 RCTs; n = 208), gestational weight gain (MD, −2.2 kg; 95% CI, −4.05, −0.34; 2 RCTs; n = 112), and excessive gestational weight gain (as defined by the National Academy of Medicine) (RR, 0.58; 95% CI, 0.35, 0.95; 1 RCT; n = 89).

Neonatal outcomes

Compared to conventional therapy, with closed‐loop pump use, there may be a reduction in the risk of stillbirth (RR, 3.13, 95% 0.13, 74.90; 2 RCTs; n = 115; low certainty). There may be little to no difference on gestational age at birth (MD, −0.37 weeks; 95% CI, −0.81, 0,08; 2 RCTs; n = 112; low certainty) and birthweight (MD, −57.32 g; 95% CI, −270.96, 156.32; 2 RCTs; n = 112; low certainty).

4.3.2. Pregnant women with T2DM

Summary of included studies

We identified three trials that compared oral agents with insulin or compared different insulin types in pregnant women with T2DM.

• One trial [36] (n = 25, USA, 2009–2011) compared metformin versus insulin.

• The MiTy trial [37] (n = 502, Australia and Canada, 2011–2018) compared metformin versus placebo; both groups continued the use of insulin.

• One trial [38] (n = 108, USA 2018–2020) compared insulin detemir versus NPH; both groups also received short‐acting insulin.

Risk of bias and research integrity assessment

The three studies had at least one outcome rated as having “some concerns” in risk of bias, primarily due to missing data. All three studies were assessed as having no or only minor concerns using the RIA tool.

Effects of interventions

Comparison 4.1 Metformin compared to insulin

The certainty of the evidence for all reported outcomes was judged to be very low.

Comparison 4.2 Metformin + insulin compared to placebo + insulin

Maternal outcomes

The combination of metformin plus insulin compared to insulin alone probably reduces the risk of caesarean birth (RR, 0.85; 95% CI, 0.73, 0.99; 1 RCT; n = 270; moderate certainty) and gestational weight gain (MD, −1.80 kg; 95% CI, −2.7, −0.9; 1 RCT; n = 482; moderate certainty), but may result in little to no effect on risk of hypertensive disorders of pregnancy (RR, 1.24; 95% CI, 0.80, 1.94; 1 RCT; n = 482 women; low certainty).

Neonatal outcomes

The combination of metformin plus insulin compared to insulin alone probably reduces the risk of macrosomia (RR, 0.63; 95% CI, 0.41, 0.97; 1 RCT; n = 461 babies; moderate certainty) but increases the risk of small‐for‐gestational age babies (RR, 1.97; 95% CI, 1.09, 3.55; 1 RCT; n = 460; moderate certainty). There is likely no difference in gestational age at birth (MD, −0.10 weeks; 95% CI, −0.48, 0.28; 1 RCT; n = 461; moderate certainty), and there is also low certainty evidence of little to no effect on the risk of admission to neonatal intensive care (RR, 1.10; 95% CI, 0.79, 1.53; 1 RCT; n = 459), preterm birth (<37 weeks) (RR, 1.37; 95% CI, 0.91, 1.77; 1 RCT; n = 461), large‐for‐gestational age (RR, 0.74; 95% CI, 0.54, 1.02; 1 RCT; n = 461), neonatal hypoglycemia (RR, 0.78; 95% CI, 0.49, 1.26; 1 RCT; n = 459), and hyperbilirubinemia (RR, 1.43; 95% CI, 0.99, 2.07; 1 RCT; n = 454).

Comparison 4.3 Insulin type: Detemir compared to NPH

Maternal outcomes

Compared to NPH, detemir may result in lower gestational weight gain (MD, −3.62 kg; 95% CI, −3.99, −3.25; 1 RCT; n = 103; low certainty).

Neonatal outcomes

Compared to NPH, detemir may increase birthweight (MD, 450.30 g; 95% CI, 163.63, 736.97; 1 RCT; n = 103; low certainty).

Health service outcomes

Compared to NPH, detemir may result in little to no difference in neonatal hospitalization (MD, 0.52 days; 95% CI 0.15, 0.89; 1 RCT; n = 103; low certainty).

4.3.3. Pregnant women with GDM

Summary of included studies

A total of 61 trials reporting relevant outcomes were identified.

Eight trials investigated initial treatment for GDM (at the point of diagnosis), comparing pharmacotherapy with no pharmacotherapy (Comparisons 5.1 and 5.2).

• One trial [39] (n = 395, USA, 2008–2012) compared glibenclamide (2.5–20 mg daily, titrated as needed) with placebo. Insulin was started if glycemic targets were not met.

• Three trials [40, 41, 42] (n = 717, Ireland, Malaysia, Thailand, 2017–2022) compared metformin (500–1000 mg daily, increased up to 2500 mg) with placebo. Two trials added insulin if glycemic targets were not met [40, 42], and one trial allowed metformin in the control group [42].

• One trial [43] (n = 263, China, 2010–2015) compared sitagliptin (100 mg daily) with placebo. Women were withdrawn from the study if glycemic targets were not met.

• Three trials [44, 45, 46] (n = 151, UK, USA, 1985–1990) compared insulin plus diet with diet and/or exercise. Insulin regimens included biphasic insulin [46], NPH + regular insulin [44]. One trial did not report the type of insulin [45].

Fifty‐two trials compared different glucose‐lowering agents or regimens (Comparisons 5.3–5.7).

• One trial [47] (n = 43, Brazil, 2003–2004) compared glibenclamide (starting dose 5 mg daily, titrated up to 20 mg if needed) with acarbose (150 mg daily, titrated up to 300 mg). Women were switched to insulin if needed.

• Three trials [48, 49, 50] (n = 510, Brazil, India, USA, 2003–2010) compared metformin (starting dose 500–1000 mg daily, titrated up to 2000–2500 mg) with glibenclamide (starting dose 2.5–5.0 mg daily, titrated up to 15–20 mg) in women who did not achieve satisfactory glycemic goals after a trial of diet/exercise. In all studies, treatment was switched to insulin or insulin was added if needed.

• Two trials [47, 51] (n = 150, Brazil 2003 and Mexico 2020) compared acarbose (75–300 mg daily) with insulin in women who did not achieve glycemic goals with diet/exercise alone. Insulin was added if needed.

• Nine trials [47, 52, 53, 54, 55, 56, 57, 58, 59] (n = 2039, Brazil, France, India, Iran, USA, 2000–2019) compared glibenclamide (starting dose 1.5–5.0 mg, titrated up to 15–20 mg) with insulin. All trials but one started pharmacotherapy after a trial of diet/exercise. Women were switched to insulin if needed, although two excluded women did not respond to pharmacotherapy.

• Twenty‐seven trials [60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86] (n = 5291, Brazil, Egypt, Finland, India, Iran, Libya, New Zealand, Pakistan, Poland, Spain, 2002–2024) compared metformin (starting dose 425–1700 mg, titrated up to 1500–3000 mg) with insulin. Most trials started pharmacotherapy after a trial of diet/exercise. Metformin was added if needed in 12 trials.

• Two trials [87, 88] (n = 352, China, 2015–2020) compared metformin plus insulin (metformin 1000 mg + insulin aspart) with insulin alone. Both studies included women who did not achieve glycemic goals with diet/exercise alone.

• One trial [89] (n = 323, India, 2008–2009) compared premixed analogue with premixed human insulin among women who had been unable to achieve blood glucose targets after a trial of diet/exercise.

• Five trials [90, 91, 92, 93, 94] (n = 330, Iran, Italy, USA, 1999–2019) compared a rapid‐acting analogue (aspart or lispro) with short‐acting insulin (regular human insulin). NPH insulin was available to all women.

• One trial [95] (n = 23, Finland, 2002) compared insulin protaphane (human intermediate‐acting) with short‐acting insulin (regular human insulin) in women unable to achieve blood glucose targets with diet/exercise.

• Two trials [96, 97] (n = 754, Israel and Pakistan, 2014–2015, one study did not report study dates) compared a twice‐daily with a four‐times‐daily insulin regimen.

Two trials evaluated specific pharmacotherapy protocols (Comparisons 5.8 and 5.9).

• One trial [98] (n = 108, Israel, 2012–2014) evaluated different pharmacotherapy sequencing—glibenclamide first, then metformin versus metformin first, then glibenclamide; insulin was given if glycemic goals were not achieved.

• One study [99] (n = 23, UK, 2014–2015) included women for whom the maximum tolerated metformin dose (up to 2000 mg) failed to achieve glycemic goals. They were assigned to glibenclamide (plus insulin if needed) or insulin.

Risk of bias and research integrity assessment

Only two trials [40, 58] were rated as having “low concerns” across all domains for all reported outcomes. All other trials had at least one outcome judged as having “some concerns” or “high concerns,” mainly due to issues with the randomization process, participant cross‐over, missing outcome data, and lack of or retrospective registration.

Of the 61 included trials, 28 were judged to have serious concerns according to the RIA assessment, primarily due to a lack of registration or retrospective registration for studies conducted after 2010. These studies were excluded from the sensitivity analysis; however, their exclusion did not significantly affect the overall results. An additional nine studies met the inclusion criteria but failed the research integrity assessment and were therefore excluded from the review. These studies not only lacked trial registration but also raised concerns regarding the conduct of the trial and the reliability of the results.

Effects of interventions

Comparison 5.1 Oral anti‐diabetic agents compared to placebo

Maternal outcomes

Moderate certainty evidence from one trial suggests that glibenclamide probably has little to no effect on risk of induction of labor (RR, 1.18; 95% CI, 0.79, 1.76; n = 375), caesarean birth (RR, 1.03; 95% CI, 0.79, 1.34; n = 375), and hypertensive disorders of pregnancy (RR, 1.24; 95% CI, 0.81, 1.90; n = 375). There is also low certainty evidence from the same trial of little to no difference on the risk of instrumental vaginal birth (RR, 0.42; 95% CI, 0.11, 1.61; n = 375), childbirth‐related injuries (RR, 0.20; 95% CI, 0.02, 1.67; n = 375), initiation of insulin (RR, 0.98; 95% CI, 0.25, 3.88; n = 375), and gestational weight gain (MD, 0.0 kg; 95% CI ‐0.97, 0.97; n = 375).

Metformin probably increases maternal satisfaction (Diabetes Treatment Satisfaction Questionnaire, RR 1.14; 95% CI, 1.01, 1.27; 1 RCT; n = 464; moderate certainty) and reduces the need for initiation of insulin (RR, 0.75, 95% 0.62, 0.91; 2 RCTs; n = 622; high certainty). However, metformin has no impact on gestational weight gain (MD, −1.13; 95% CI, −1.63, −0.64; n = 3, n = 705; high certainty), probably has little to no effect on risk of induction of labor (RR, 0.83; 95% CI, 0.65, 1.06; 2 RCTs; n = 629; moderate certainty), caesarean birth (RR, 1.10; 95% CI, 0.91, 1.32; 3 RCTs; n = 704; moderate certainty), and hypertensive disorders of pregnancy (RR, 1.26; 95% CI, 0.84, 1.89; 3 RCTs; n = 693; moderate certainty), and may also have little to no effect on risk of instrumental vaginal birth (RR, 0.67; 95% CI, 0.22, 2.02; 2 RCTs; n = 180; low certainty).

Neonatal outcomes

Glibenclamide may have little to no effect on risk of stillbirth/fetal death (RR, 0.49; 95% CI, 0.05, 5.38; 1 RCT; n = 375; low certainty). There is moderate certainty evidence from the same trial of little to no effect on birthweight (MD, −33.00; 95% CI, 134.53, 68.53; n = 375) and risk of small‐for‐gestational age (RR, 1.11; 95% CI, 0.58, 2.10; n = 375), large‐for‐gestational age (RR, 0.89; 95% CI, 0.51, 1.58; n = 375) and macrosomia (RR, 0.71; 95% CI, 0.36, 1.41; n = 375), and low certainty evidence of little to no effect on admission to neonatal intensive care (RR, 1.16; 95% CI, 0.53, 2.53; n = 375), gestational age at birth (MD, 0 weeks; 95% CI, −0.32, 0.32 higher; n = 375), shoulder dystocia (RR, 0.33; 95% CI, 0.001, 8.00; n = 375), and hyperbilirubinemia (RR, 1.97; 95% CI, 0.50, 7.75; n = 375).

Metformin reduces birthweight (MD, −137.58 g; 95% CI, 207.23, −67.92; 3 RCTs; n = 705; high‐certainty) and the risk of large‐for‐gestational age (RR, 0.44; 95% CI, 0.25, 0.75; 1 RCT; n = 522; high certainty) but increases the risk of low birthweight (<2550 g) (RR, 1.93; 95% CI, 1.04, 3.58; 2 RCTs; n = 628; high‐certainty). However, there is moderate certainty evidence of little to no difference on risk of admission to neonatal intensive care (RR, 1.23; 95% CI, 0.85, 1.79; 3 RCTs; n = 700), preterm birth (RR, 1.41; 95% CI, 0.78, 2.57; 1 RCT; n = 524 babies), gestational age at birth (MD, −1.57 weeks; 95% CI, 5.41, 2.26; 2 RCTs; n = 599), macrosomia (RR, 0.79; 95% CI, 0.15, 4.03; 3 RCTs; n = 705), neonatal hypoglycemia (RR, 1.08; 95% CI, 0.70, 1.65; 3 RCTs; n = 700), and respiratory distress syndrome (RR, 1.44; 95% CI, 0.84, 2.45; 3 RCTs; n = 700); and low certainty evidence of little to no effect on the risk of neonatal death (RR, 0.33; 95% CI, 0.01, 8.08; 1 RCT; n = 526), small‐for‐gestational age (RR, 2.14; 95% CI, 0.89, 5.17; 1 RCT; n = 522), Apgar score <7 at 5 min (RR, 0.99; 95% CI, 0.06,15.78; 2 RCTs; n = 597), hyperbilirubinemia (RR, 1.14; 95% CI, 0.45, 2.90; 3 RCTs; n = 700), transient tachypnoea of the newborn (RR, 0.34; 95% CI, 0.04, 3.16; 1 RCT; n = 101), and congenital anomalies (RR, 1.43; 95% CI, 0.55, 3.70; 1 RCT; n = 524).

Health service outcomes

Metformin may have little to no effect on length of neonatal hospitalization (MD, −0.1; 95% CI, −0.9, 0.7; 1 RCT; n = 75; low certainty).

Comparison 5.2 Insulin compared to diet and/or exercise

Maternal outcomes

Insulin versus diet ± exercise may increase gestational weight gain in the first month of treatment (MD, 2.6 kg; 95% CI, 0.06, 5.14; 1 RCT; n = 15; low certainty).

Comparison 5.3 Type of oral anti‐diabetic agents: Glibenclamide compared to acarbose

Maternal outcomes

Glibenclamide versus acarbose may have little to no effect on gestational weight gain (MD, −0.60 kg; 95% CI, −3.13, 1.93; 1 RCT; n = 43; low certainty).

Neonatal outcomes

Glibenclamide versus acarbose may have little to no effect on gestational age at birth (MD, −0.10 weeks; 95% CI, 0.82, 0.62; 1 RCT; n = 43; low certainty).

Comparison 5.4 Type of oral anti‐diabetic agents: Metformin compared to glibenclamide

Maternal outcomes

Metformin versus glibenclamide may reduce gestational weight gain (MD, −2.06 kg; 95% CI, −3.98, 0.14; 1 RCT; n = 200; low certainty), but may have little to no effect on risk of caesarean birth (RR, 1.03; 95% CI, 0.83, 1.29; 3 RCTs; n = 508; low certainty), hypertensive disorders of pregnancy (RR, 0.71; 95% CI, 0.38, 1.31; 3 RCTs; n = 508; low certainty), and need for initiation of insulin (RR, 1.23; 95% CI, 0.43, 3.53; 2 RCTs; n = 349; low certainty).

Neonatal outcomes

Metformin versus glibenclamide may reduce the risk of large‐for‐gestation age (RR, 0.44; 95% CI, 0.21 to 0.92; 1 trial, 200 babies; low certainty), but may may have little to no effect on risk of stillbirth/fetal death (RR, 0.92; 95% CI, 0.06, 14.55; 2 RCTs; n = 359; low certainty), perinatal death (RR, 0.92; 95% CI, 0.06, 14.55; 1 RCT; n = 200; low certainty), gestational age at birth (MD, −0.11 weeks; 95% CI, −0.40, 0.19; 2 RCTs; n = 349; low certainty), birthweight (MD, −118.55 g; 95% CI, −284.83, 47.74; 3 RCTs; n = 508; low certainty), and risk of hypoglycemia (RR, 0.52; 95% CI, 0.08, 3.29; 3 RCTs; n = 508; low certainty).

Comparison 5.5 Oral anti‐diabetic agent compared to insulin

Maternal outcomes

Glibenclamide versus insulin probably increases the risk of hypoglycemia (RR, 3.95; 95% CI, 1.33, 11.71; 3 RCTs; n = 988; moderate certainty), but probably has little to no effect on risk of caesarean birth (RR, 0.97; 95% CI, 0.84, 1.11; 4 RCTs; n = 1108; moderate certainty), instrumental vaginal birth (RR, 1.07; 95% CI, 0.79, 1.46; 1 RCT; n = 890; moderate certainty), gestational weight gain (MD, 0.21 kg; 95% CI, −0.87, 1.30; 3 RCTs; n = 526; moderate certainty), and may also have little to no effect on the risk of childbirth‐related injuries (RR, 2.96; 95% CI, 0.31, 28.35; 2 RCTs; n = 1139; low certainty).

Metformin versus insulin probably has little to no effect on the risk of induction of labor (RR, 0.85; 95% CI, 0.71, 1.01; 4 RCTs; n = 1237; sensitivity analysis; moderate certainty), and may also have little to no effect on the risk of caesarean birth (RR, 0.94; 95% CI, 0.88, .01; 18 RCTs; n = 4055; low certainty), instrumental vaginal birth (RR, 1.02; 95% CI, 0.61, 1.69; 5 RCTs; n = 730; low certainty), hypertensive disorders of pregnancy (RR, 0.76; 95% CI, 0.57, 1.00; 10 RCTs; n = 2830; low certainty), and gestational weight gain (MD, −1.44 kg; 95% CI, −2.16, −0.73; 9 RCTs; n = 1740; low certainty).

Neonatal outcomes

Glibenclamide versus insulin increases the risk of macrosomia (RR, 1.57; 95% CI, 1.07, 2.32; 5 RCTs; n = 1478; sensitivity analysis; high certainty) and probably increases the risk of neonatal hypoglycemia (RR, 1.46; 95% CI, 1.06, 2.02; 9 RCTs; n = 1967; moderate certainty), but there is high certainty evidence of no difference in birthweight (MD, 65.04 g; 95% CI, −2.22, 132.30; 7 RCTs; n = 1585; sensitivity analysis), moderate certainty evidence of probably no difference in the risk of preterm birth (RR, 1.34; 95% CI, 0.84, 2.14; 4 RCTs; n = 1130 babies), admission to neonatal intensive care (RR, 1.04; 95% CI, 0.75, 1.46; 6 RCTs; n = 1562; sensitivity analysis); gestational age at birth (MD, 0.1 weeks; 95% CI, −0.15, 0.35; 4 RCTs; n = 588; sensitivity analysis) and hyperbilirubinemia (RR, 1.45; 95% CI, 0.87, 2.42; 4 RCTs; n = 1847; sensitivity analysis), and low certainty evidence of little to no difference in risk of neonatal death (RR, 1.01; 95% CI, 0.06, 16.04; 2 RCTs; n = 486 babies), perinatal death (RR0.20; 95% CI, 0.01, 4.12; 2 RCTs; n = 930), birth injury (RR, 0.77; 95% CI, 0.29, 2.04; 2 RCTs; n = 928), Apgar score <7 at 5 min (RR, 0.46; 95% CI, 0.17, 1.23; 2 RCTs; n = 986), and respiratory distress syndrome (RR, 0.73; 95% CI, 0.40, 1.34; 5 RCTs; n = 1723).

Metformin versus insulin probably reduces the risk of admission to neonatal intensive care (RR, 0.75; 95% CI, 0.60, 0.93; 16 RCTs; n = 3184; moderate certainty) and of neonatal hypoglycemia (RR, 0.77; 95% CI, 0.61, 0.97; 6 RCTs; n = 1524; sensitivity analysis; moderate certainty), but it likely has little to no effect on risk of gestational age at birth (MD, −0.15 weeks; 95% CI, −0.31, 0.01; 4 RCTs; n = 1233; sensitivity analysis; moderate certainty), and birthweight (MD, −21.84 g; 95% CI, ‐75.81, 32.13; 6 RCTs; n = 1529; sensitivity analysis; moderate certainty). There is also low certainty evidence of little to no difference on preterm birth (RR, 1.23; 95% CI, 0.83, 1.80; n = 1533; sensitivity analysis), small‐for‐gestational age (RR, 0.92; 95% CI, 0.65, 1.31; 9 RCTs; n = 2002), large‐for‐gestational age (RR, 0.82; 95% CI, 0.68,1.01; 10 RCTs; n = 2215), macrosomia (RR, 0.90; 95% CI, 0.58, 1.39; 5 RCTs; n = 796; sensitivity analysis), respiratory distress syndrome (RR, 0.68; 95% CI, 0.43, 1.05; 4 RCTs; n = 1213; sensitivity analysis), birth injury (RR, 1.12; 95% CI, 0.69, 1.82; 4 RCTs; n = 1233), hyperbilirubinemia (RR, 0.80; 95% CI, 0.59, 1.09; n = 1522; sensitivity analysis), and congenital anomalies (RR, 0.67; 95% CI, 0.40, 1.14; 8 RCTs; n = 1816).

Concerning long‐term outcomes, metformin probably has little to no effect on child cardiometabolic outcomes at 24 months (weight: MD 0.30 kg; 95% CI, −0.17, 0.77 kg; 1 RCT; n = 318 children; moderate certainty), and it may have little to no effect on development at 18 months (standing without support: MD ‐0.1 month; 95% CI, −0.91, 0.71; and walking without support: MD 0.3 months; 95% CI, −0.56, 1.16; 1 RCT; n = 93; low certainty) and on neurodevelopmental stage at 24 months (Bayley Scales of Infant Development, Second Edition [BSID II] mental development index: MD ‐0.83; 95% CI, −5.6, 3.94; and psychomotor development index: MD 0.62; 95% CI ‐4.00, 5.23; 1 RCT; n = 211; moderate certainty). Another trial also reports similar results across all Bailey III scale domains.

Changes from the main analysis with sensitivity analysis

For glibenclamide compared to insulin, certainty increased from very low to high for macrosomia, from low to high for birthweight, and from low to moderate for admission to neonatal intensive care, gestational age at birth, and hyperbilirubinemia. Certainty decreased from low to very low for large‐for‐gestational age (see Table 3).

TABLE 3.

Summary of changes in conclusion and/or certainty with sensitivity analysis.

Outcome	Main analysis	Sensitivity analysis
Glibenclamide compared to insulin
Macrosomia	RR, 1.25 (95% CI, 0.56, 2.82) Very low (⨁◯◯◯)	RR, 1.57 (95% CI, 1.07, 2.32) High (⨁⨁⨁⨁)
Birthweight	MD, 29.56 g (95% CI, −79.35, 138.46) Low (⨁⨁◯◯)	MD, 65.04 g (95% CI, −2.22, 132.30) High (⨁⨁⨁⨁)
Admission to neonatal intensive care	RR, 0.99 (95% CI, 0.72, 1.35) Low (⨁⨁◯◯)	RR, 1.04 (95% CI, 0.75, 1.46) Moderate (⨁⨁⨁◯)
Gestational age at birth	MD, 0.23 weeks (95% CI, −0.3, 0.75) Low (⨁⨁◯◯)	MD, 0.1 weeks (95% CI, −0.15, 0.75) Moderate (⨁⨁⨁◯)
Hyperbilirubinemia	RR, 1.08 (95% CI, 0.80, 1.66) Low (⨁⨁◯◯)	RR, 1.45 (95%CI, 0.87. 2.42) Moderate (⨁⨁⨁◯)
Large‐for‐gestational age	RR, 1.86 (95% CI, 0.79, 4.37) Low (⨁⨁◯◯)	RR, 1.94 (95% CI, 0.65, 5.76) Very low (⨁◯◯◯)
Metformin compared to insulin
Induction of labor	RR, 0.84 (95% CI, 0.73, 0.98) Moderate (⨁⨁⨁◯)	RR, 0.85 (95% CI, 0.71, 1.01) Moderate (⨁⨁⨁◯)
Macrosomia	RR, 0.67 (95% CI, 0.51, 0.87) Low (⨁⨁◯◯)	RR, 0.90 (95% CI, 0.58 to 1.39) Low (⨁⨁◯◯)
Respiratory distress syndrome	RR, 0.69 (95% CI, 0.47, 0.99) Low (⨁⨁◯◯)	RR, 0.68 (95% CI, 0.43, 1.05) Low (⨁⨁◯◯)
Neonatal hypoglycemia	RR, 0.59 (95% CI, 0.49, 0.71) Low (⨁⨁◯◯)	RR, 0.77 (95% CI, 0.61, 0.97) Moderate (⨁⨁⨁◯)
Gestational age at birth	MD, 0.05 weeks (95% CI, −0.2, 0.3) Low (⨁⨁◯◯)	MD, −0.15 (95% CI, −0.31, 0.01) Moderate (⨁⨁⨁◯)
Birthweight	MD, −79.5 g (‐153.13, −5.87) Very low (⨁◯◯◯)	MD, −21.84 g (95% CI, −75.81, 32.13) Moderate (⨁⨁⨁◯)
Preterm birth	RR, 0.83 (95% CI, 0.60, 1.16) Very low (⨁◯◯◯)	RR, 1.23 (95% CI, 0.83, 1.80) Low (⨁⨁◯◯)
Hyperbilirubinemia	RR, 0.86 (95% CI, 0.64, 1.14) Very low (⨁◯◯◯)	RR, 0.80 (95% CI, 0.59, 1.09) Low (⨁⨁◯◯)
Hypoglycemia	RR, 0.43 (95% CI, 0.26, 0.71) Low (⨁⨁◯◯)	RR, 0.32 (95%CI, 0.20 to 0.51) Very low (⨁◯◯◯)
Shoulder dystocia	RR, 0.41 (95% CI, 0.17, 0.99) Low (⨁⨁◯◯)	RR, 0.48 (95% CI, 0.10, 2.33) Very low (⨁◯◯◯)
Caesarean birth	RR, 0.94 (95% CI, 0.88, 1.01) Low (⨁⨁◯◯)	RR, 0.93 (95% CI, 0.65 to 1.35) Very low (⨁◯◯◯)

For metformin compared to insulin, there was a change from a reduction of risk to little to no difference in effect for induction of labor, macrosomia, and respiratory distress syndrome, but no change in certainty. There was an increase in certainty from low to moderate for neonatal hypoglycemia and gestational age at birth, from very low to moderate for birthweight, from very low to low for preterm birth and hyperbilirubinemia. There was a reduction in certainty from low to very low for maternal hypoglycemia, shoulder dystocia, and caesarean birth (see Table 3).

Comparison 5.6 Type of insulin

Maternal outcomes

Premixed analogue versus premixed human insulin probably has little to no effect on risk of caesarean birth (RR, 0.98; 95% CI, 0.91, 1.06; 1 RCT; n = 320; moderate certainty) and may have little to no effect on risk of hypoglycemia (RR, 1.19; 95% CI, 0.65, 2.17; 1 RCT; n = 320; low certainty) and gestational weight gain (MD, −0.47 kg; 95% CI, −1.09, 0.15; 1 RCT; n = 320; low certainty).

Rapid‐acting versus short‐acting insulin may result in little to no difference in the risk of hyperglycemia (MD, −1.5% percentage of determinations in hyperglycemic range; 95% CI, −2.83, −0.17; 1 RCT; n = 320; low certainty).

Neonatal outcomes

Premixed analogue versus premixed human insulin may have little to no effect on gestational age at birth (MD, 0.67 weeks; 95% CI, 0.33,1.01; 1 RCT; n = 320; low certainty), birthweight (MD, 10.0 g; 95% CI, −93.79, 113.79; 1 RCT; n = 320; low certainty) or risk of small‐for‐gestational age (RR, 0.86; 95% CI, 0.47, 1.60; 1 RCT; n = 320; low certainty).

Rapid‐acting versus short‐acting insulin may have little to no effect on gestational age at birth (MD, 0 weeks; 95% CI, −0.71, 0.71; 1 RCT; n = 41; low certainty) and birthweight (MD, 42.42 g; 95% CI, −111.26, 196.1; 3 RCTs; n = 117; low certainty).

Comparison 5.7 Insulin regimen: twice daily compared to four times daily insulin

Maternal outcomes

Insulin twice daily versus four times daily may decrease the risk of hypoglycemia (RR, 0.42; 95% CI, 0.35 to 0.51; 2 RCTs; n = 754; low certainty).

Neonatal outcomes

Insulin twice daily versus four times daily may increase the risk of neonatal hypoglycemia (RR, 8.50; 95% CI, 5.23, 13.82; 1 RCT; n = 480; low certainty) and hyperbilirubinemia (RR, 3.40; 95% CI, 2.51, 4.61; 1 RCT; n = 480; low certainty).

Comparison 5.8 Glibenclamide (1st line), Metformin (2nd line), Insulin (3rd line) compared to Metformin (1st lone), Glibenclamide (2nd line), Insulin (3rd line)

Maternal outcomes

A protocol of glibenclamide first, then metformin versus metformin first, then glibenclamide (followed by insulin if needed) may have little to no effect on the need for insulin as third‐line treatment (RR, 4.33; 95% CI 0.98, 19.08; 1 RCT; n = 104; low certainty) or on gestational weight gain (MD, 0.30 kg; 95% CI, −2.32, 2.92; 1 RCT; n = 104; low certainty).

Neonatal outcomes

A protocol of glibenclamide first, then metformin versus metformin first, then glibenclamide (followed by insulin if needed) may have little to no effect on gestational age at birth (MD, 0.5 weeks; 95% CI, −0.02, 1.02; 1 RCT; n = 104; low certainty) and birthweight (MD, −0.50 g; 95% CI, −239.14, 139.14; 1 RCT; n = 104; low certainty).

Comparison 5.9 Metformin (1st line), glibenclamide (2nd line) compared to metformin (1st line), insulin (2nd line)

Maternal outcomes

Low certainty evidence from one trial suggests that in women for whom metformin fails to achieve adequate glycemic stability, glibenclamide compared to insulin may have little to no effect on the risk of caesarean birth (RR, 0.51; 95% CI, 0.20, 1.43; n = 23), instrumental vaginal birth (RR, 0.77; 95% CI, 0.05, 10.85; n = 23), and gestational weight gain (MD, −0.30 kg lower; 95% CI, −4.41, 3.81; n = 20).

Neonatal outcomes

Low certainty evidence from one trial suggests that in women for whom metformin fails to achieve adequate glycemic stability, glibenclamide compared to insulin may increase birthweight (MD, 444 g; 95% CI, 16.62, 871.38; n = 23), but may have little to no effect on risk of preterm birth (RR, 0.11; 95% CI, 0.01, 1.95; n = 23), admission to neonatal intensive care (RR, 3.08; 95% CI, 0.40, 23.44; n = 23), birth injury (RR, 0.26; 95% CI, 0.01, 5.82; n = 23), neonatal hypoglycemia (RR, 2.45; 95% CI, 0.31, 19.74; n = 20), hyperbilirubinemia (RR, 0.38; 95% CI, 0.04, 3.67; n = 23), transient tachypnoea of the newborn (RR, 2.36; 95% CI, 0.11, 52.41; n = 23), and congenital anomalies (RR, 0.26; 95% CI, 0.01, 5.82; n = 23).

5. DISCUSSION

5.1. Summary of main results

This systematic review provides a comprehensive synthesis of the evidence on pharmacological glucose‐lowering interventions for the management of diabetes in pregnancy and was used to support the development of the new WHO guidelines on the management of diabetes in pregnant women.

5.1.1. Glycemic targets in pregnancy

The evidence on glycemic targets in pregnancy is limited, with most outcomes rated as low or very low certainty. In women with T1DM, tighter glycemic targets may increase the risk of hypoglycemia and longer postnatal hospitalization, whereas looser control may increase the risk of caesarean birth, large‐for‐gestational‐age babies, and neonatal respiratory distress syndrome. Among women with GDM, there were no clear differences in maternal or neonatal outcomes when comparing tight versus moderate glycemic targets, except for a small increase in antenatal hospital stay associated with a tighter control. No evidence was identified for women with T2DM.

5.1.2. Pharmacological glucose‐lowering interventions in women with T1DM

Across 17 trials evaluating approaches to pharmacotherapy in women with T1DM, rapid‐acting insulin may result in increased satisfaction and reduced risk of preterm birth but may have no effect on hypertensive disorders of pregnancy and gestational age at birth. Long‐acting insulin compared to NPH may increase the risk of preeclampsia but probably results in little to no difference in risk of major hypoglycemia and may also result in little to no difference in hypoglycemia, gestational age at birth, and birthweight. When degludec is compared to detemir, there is possibly little to no difference in maternal and neonatal outcomes.

Evidence comparing different insulin delivery methods and timings, including CSII versus conventional insulin, sensor‐augmented insulin pump in predictive low glucose suspend (PLGS) mode versus sensor‐augmented insulin pump in low glucose suspend (LGS), and insulin given before versus after meals was of very low certainty. Closed‐loop pump compared to conventional therapy may result in a reduction in the risk of stillbirth, caesarean section, gestational hypertension, gestational weight gain, but little or no difference in gestational age at birth and birthweight.

5.1.3. Pharmacological glucose‐lowering interventions in women with T2DM

Only three trials evaluated pharmacotherapy in women with T2DM. The evidence on metformin versus insulin was too uncertain to draw any conclusions. The combination of metformin‐insulin versus insulin alone probably reduces the risk of caesarean birth, gestational weight gain, and macrosomia but probably increases the risk of small‐for‐gestational‐age babies. Detemir compared to NPH insulin may result in lower gestational weight gain, but increase birthweight, and little to no difference in neonatal hospital stay.

5.1.4. Pharmacological glucose‐lowering interventions in women with GDM

Eight trials compared a glucose‐lowering agent versus placebo and/or usual care as initial treatment. The evidence on sitagliptin versus placebo (one trial) and insulin versus no insulin (three trials) was too sparse to support firm conclusions. Glibenclamide compared to placebo (one trial) showed very low to moderate certainty evidence of little to no benefit for most reported maternal and neonatal outcomes. When comparing metformin to placebo (three trials), there was high certainty evidence of reduced need for insulin initiation, slightly reduced birthweight and reduced risk of large‐for‐gestational‐age babies but increased the risk of low birthweight. There was also moderate certainty evidence of increased maternal satisfaction. For all other reported maternal and neonatal outcomes, there was very low to high certainty evidence of little to no difference.

Four trials compared different oral glucose‐lowering agents as first‐line treatment. Metformin compared with glibenclamide (three trials) may reduce gestational weight gain and the risk of large‐for‐gestational‐age babies but likely has little to no difference on all other maternal and neonatal outcomes. The evidence on glibenclamide versus acarbose (one trial) is very limited for meaningful interpretation.

Thirty‐seven trials compared oral glucose‐lowering agents with insulin as first‐line treatment. Metformin compared to insulin (27 trials) probably reduces the risk of neonatal intensive care admission, neonatal hypoglycemia, and may reduce gestational weight gain. There is very low to moderate certainty evidence of little to no effect on other neonatal, maternal, and long‐term child outcomes. When comparing glibenclamide versus insulin (nine trials), moderate to high certainty evidence suggests that it probably increases the risk of macrosomia and neonatal and maternal hypoglycemia; however, there is low to moderate certainty evidence of little to no effect on other neonatal and maternal outcomes. The data on acarbose versus insulin (two trials) are too sparse to draw reliable inferences. In addition, two trials evaluated the combination of metformin and insulin versus insulin alone, but the evidence base is also very limited.

Nine studies compared different types of insulin or insulin regimens, but the findings are inconclusive due to limited data (many outcomes were not reported or assessed as very low certainty).

Two studies evaluated specific treatment protocols. In one study, women requiring escalation therapy after inadequate control with metformin were randomized to glibenclamide or insulin. Glibenclamide may increase birthweight but may have little to no effect on other neonatal or maternal outcomes. Another study compared two protocols, one initiating with glibenclamide followed by metformin and insulin, and the other starting with metformin followed by glibenclamide and insulin, found little to no difference in maternal and neonatal outcomes, but many outcomes were rated as very low certainty or not reported.

5.2. Agreements and disagreements with other reviews

Two Cochrane reviews [100, 101] have previously examined glycemic targets in pregnancy and similarly reported a limited evidence base. Middleton et al. [101] evaluated the effects of different intensities of glycemic goals in pregnant women with pre‐existing diabetes and included the same three trials as our review. The review concluded that “loose” targets may result in harm (increased risk of preeclampsia, caesarean birth, and large‐for‐gestational‐age babies), whereas a “very tight” target may result in more hypoglycemia episodes and longer hospitalization. Hofer et al. [100] focused on women with GDM and included the same three trials as our review, along with another study published as a conference abstract [102]. They concluded that tighter glycemic targets may increase the risk of hypertensive disorder in pregnancy.

Several have evaluated pharmacotherapy for the management of diabetes in pregnancy. One Cochrane review [103] including six RCTs involving women with pre‐existing diabetes, impaired glucose tolerance, previous gestational diabetes, or gestational diabetes concluded that there is insufficient evidence to evaluate the use of oral agents in these populations. Another Cochrane review [104] evaluated different types of insulin and insulin regimens in pregnant women with pre‐existing T1DM or T2DM, and also concluded that the evidence was too limited to make firm conclusions. A third Cochrane review [105] evaluated the use of CSII versus multiple daily injections in pregnant women with T1DM, including four trials also captured in our review and an additional conference abstract [106]. The review also concluded that there is limited and dated evidence to recommend one mode of delivery over another. Another review [107] including RCTs and non‐randomized studies found that CSII achieved better glycemic stability during the first trimester, but this advantage diminished later in pregnancy. CSII was also associated with lower insulin requirements, greater gestational weight gain, and a differing risk of large‐ and small‐for‐gestational‐age babies. A more recent meta‐analysis [108] including 39 randomized and non‐randomized studies found that CSII was associated with higher odds of caesarean birth and large‐for‐gestational age babies.

Brown et al. [109] evaluated the evidence on oral agents compared to placebo or non‐pharmacological interventions in GDM. The review included 11 trials, and the authors concluded that there was insufficient evidence to determine the benefits of oral agents compared to placebo or standard care. Another review by the same author [110] evaluated the use of insulin in women with GDM. Data were limited for the comparison of insulin versus standard care. Compared to oral agents, insulin was associated with a higher risk of hypertensive disorders of pregnancy and induction of labor, although the evidence was uncertain. Among the 10 trials comparing different types of insulin, there was no evidence of an overall difference between the groups for any of the primary or secondary maternal or neonatal outcomes. Similarly, for the trials comparing insulin versus standard care, data were poorly reported.

A network meta‐analysis [111] evaluated the efficacy of metformin, glibenclamide, and insulin as first‐line treatment for the management of GDM. The authors concluded that metformin was comparable to insulin for the treatment of GDM, while glibenclamide appeared to be less favorable. Another network meta‐analysis [112] also comparing different glucose‐lowering treatments in women with GDM reached a similar conclusion. Metformin appeared to be superior to other options, as it was associated with fewer adverse outcomes compared to glibenclamide or the combination metformin–insulin.

5.3. Quality of the evidence

The certainty of the evidence was generally low to very low, with only a few outcomes graded as moderate or high. Downgrading was most often due to risk of bias and imprecision. Many studies included small sample sizes, and several analyses were underpowered to detect differences between the groups. Methodological limitations were common; many of the trials were unregistered or registered retrospectively, and study methods (randomization, allocation concealment) were often poorly described. Long‐term maternal and neonatal outcomes were poorly reported for all comparisons. All trials were assessed for research integrity, and study authors were contacted for clarification when concerns were identified. Sensitivity analyses excluding trials with unresolved serious concerns did not substantially affect the conclusions.

6. CONCLUSION

The available evidence from this review was insufficient to establish the optimal glycemic targets for improving maternal and neonatal outcomes in women with pre‐existing diabetes or with GDM.

Regarding pharmacological management, most evidence relates to GDM. For women with GDM in whom diet and/or exercise alone were inadequate, both metformin and insulin appear more effective and safe than other glucose‐lowering alternatives, although no clear superiority between the two was observed. Switching to or adding an alternative treatment is often done in studies when glycemic targets are not achieved, but evidence on the best treatment protocol was limited. The evidence on optimal insulin regimens for GDM also remains unclear. For women with T1DM, data on different insulin types and regimens are too scarce to support firm conclusions. Among women with pre‐existing T2DM who were unable to achieve adequate glycemic goals or who were already receiving insulin, there is not enough evidence to recommend one treatment over another.

AUTHOR CONTRIBUTIONS

Gemma Villanueva, Nicholas Henschke, Elise Cogo, and Yanina Sguassero conceived the study and designed the protocol. Elise Cogo developed the search strategy. Gemma Villanueva and Nicholas Henschke analyzed the data. Gemma Villanueva, Elise Cogo, and Nicholas Henschke contributed to the interpretation of the findings. Gemma Villanueva, Jennifer Petkovic, and Katrin Probyn wrote the first draft of the paper. All authors contributed to screening, data extraction, and risk of bias assessment. All authors critically revised the manuscript for intellectual content and approved the final manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS STATEMENT

Not applicable.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Supporting Information