Oral anti‐diabetic pharmacological therapies for the treatment of women with gestational diabetes

Abstract

Background

Gestational diabetes mellitus (GDM) is a major public health issue with rates increasing globally. Gestational diabetes, glucose intolerance first recognised during pregnancy, usually resolves after birth and is associated with short‐ and long‐term complications for the mother and her infant. Treatment options can include oral anti‐diabetic pharmacological therapies.

Objectives

To evaluate the effects of oral anti‐diabetic pharmacological therapies for treating women with GDM.

Search methods

We searched Cochrane Pregnancy and Childbirth's Trials Register (14 May 2016), ClinicalTrials.gov, WHO ICTRP (14 May 2016) and reference lists of retrieved studies.

Selection criteria

We included published and unpublished randomised controlled trials assessing the effects of oral anti‐diabetic pharmacological therapies for treating pregnant women with GDM. We included studies comparing oral anti‐diabetic pharmacological therapies with 1) placebo/standard care, 2) another oral anti‐diabetic pharmacological therapy, 3) combined oral anti‐diabetic pharmacological therapies. Trials using insulin as the comparator were excluded as they are the subject of a separate Cochrane systematic review.

Women with pre‐existing type 1 or type 2 diabetes were excluded.

Data collection and analysis

Two review authors independently assessed trials for inclusion and trial quality. Two review authors independently extracted data and data were checked for accuracy.

Main results

We included 11 studies (19 publications) (1487 women and their babies). Eight studies had data that could be included in meta‐analyses. Studies were conducted in Brazil, India, Israel, UK, South Africa and USA. The studies varied in diagnostic criteria and treatment targets for glycaemic control for GDM. The overall risk of bias was 'unclear' due to inadequate reporting of methodology. Using GRADE the quality of the evidence ranged from moderate to very low quality. Evidence was downgraded for risk of bias (reporting bias, lack of blinding), inconsistency, indirectness, imprecision and for oral anti‐diabetic therapy versus placebo for generalisability.

Oral anti‐diabetic pharmacological therapies versus placebo/standard care

There was no evidence of a difference between glibenclamide and placebo groups for hypertensive disorders of pregnancy (risk ratio (RR) 1.24, 95% confidence interval (CI) 0.81 to 1.90; one study, 375 women, very low‐quality evidence), birth by caesarean section (RR 1.03, 95% CI 0.79 to 1.34; one study, 375 women, very low‐quality evidence), perineal trauma (RR 0.98, 95% CI 0.06 to 15.62; one study, 375 women, very low‐quality evidence) or induction of labour (RR 1.18, 95% CI 0.79 to 1.76; one study, 375 women; very low‐quality evidence). No data were reported for development of type 2 diabetes or other pre‐specified GRADE maternal outcomes (return to pre‐pregnancy weight, postnatal depression). For the infant, there was no evidence of a difference in the risk of being born large‐for‐gestational age (LGA) between infants whose mothers had been treated with glibenclamide and those in the placebo group (RR 0.89, 95% CI 0.51 to 1.58; one study, 375, low‐quality evidence). No data were reported for other infant primary or GRADE outcomes (perinatal mortality, death or serious morbidity composite, neurosensory disability in later childhood, neonatal hypoglycaemia, adiposity, diabetes).

Metformin versus glibenclamide

There was no evidence of a difference between metformin‐ and glibenclamide‐treated groups for the risk of hypertensive disorders of pregnancy (RR 0.70, 95% CI 0.38 to 1.30; three studies, 508 women, moderate‐quality evidence), birth by caesarean section (average RR 1.20, 95% CI 1.20; 95% CI 0.83 to 1.72, four studies, 554 women, I² = 61%, Tau² = 0.07 low‐quality evidence), induction of labour (0.81, 95% CI 0.61 to 1.07; one study, 159 women; low‐quality evidence) or perineal trauma (RR 1.67, 95% CI 0.22 to 12.52; two studies, 158 women; low‐quality evidence). No data were reported for development of type 2 diabetes or other pre‐specified GRADE maternal outcomes (return to pre‐pregnancy weight, postnatal depression). For the infant there was no evidence of a difference between the metformin‐ and glibenclamide‐exposed groups for the risk of being born LGA (average RR 0.67, 95% CI 0.24 to 1.83; two studies, 246 infants, I² = 54%, Tau² = 0.30 low‐quality evidence). Metformin was associated with a decrease in a death or serious morbidity composite (RR 0.54, 95% CI 0.31 to 0.94; one study, 159 infants, low‐quality evidence). There was no clear difference between groups for neonatal hypoglycaemia (RR 0.86, 95% CI 0.42 to 1.77; four studies, 554 infants, low‐quality evidence) or perinatal mortality (RR 0.92, 95% CI 0.06 to 14.55, two studies, 359 infants). No data were reported for neurosensory disability in later childhood or for adiposity or diabetes.

Glibenclamide versus acarbose

There was no evidence of a difference between glibenclamide and acarbose from one study (43 women) for any of their maternal or infant primary outcomes (caesarean section, RR 0.95, 95% CI 0.53 to 1.70; low‐quality evidence; perinatal mortality ‐ no events; low‐quality evidence; LGA , RR 2.38, 95% CI 0.54 to 10.46; low‐quality evidence). There was no evidence of a difference between glibenclamide and acarbose for neonatal hypoglycaemia (RR 6.33, 95% CI 0.87 to 46.32; low‐quality evidence). There were no data reported for other pre‐specified GRADE or primary maternal outcomes (hypertensive disorders of pregnancy, development of type 2 diabetes, perineal trauma, return to pre‐pregnancy weight, postnatal depression, induction of labour) or neonatal outcomes (death or serious morbidity composite, adiposity or diabetes).

Authors' conclusions

There were insufficient data comparing oral anti‐diabetic pharmacological therapies with placebo/standard care (lifestyle advice) to inform clinical practice. There was insufficient high‐quality evidence to be able to draw any meaningful conclusions as to the benefits of one oral anti‐diabetic pharmacological therapy over another due to limited reporting of data for the primary and secondary outcomes in this review. Short‐ and long‐term clinical outcomes for this review were inadequately reported or not reported. Current choice of oral anti‐diabetic pharmacological therapy appears to be based on clinical preference, availability and national clinical practice guidelines.

The benefits and potential harms of one oral anti‐diabetic pharmacological therapy compared with another, or compared with placebo/standard care remains unclear and requires further research. Future trials should attempt to report on the core outcomes suggested in this review, in particular long‐term outcomes for the woman and the infant that have been poorly reported to date, women's experiences and cost benefit.

Plain language summary

Oral medication for the treatment of women with gestational diabetes

What is the issue?

Globally the number of women being diagnosed with gestational diabetes mellitus (GDM) is increasing. GDM is an intolerance to glucose leading to high blood sugars, first recognised during pregnancy and usually resolving after birth. Standard care involves lifestyle advice on diet and exercise. Treatment for some women includes oral anti‐diabetic medications, such as metformin and glibenclamide, which are an alternative to, or can be used alongside, insulin to control the blood sugar. This review aimed to investigate benefits of taking oral medication to treat GDM in pregnant women. Another Cochrane Review compares the effects of insulin with oral anti‐diabetic pharmacological therapies (Brown 2016).

Why is this important?

Women diagnosed with GDM are at a greater risk of experiencing complications such as high blood pressure during pregnancy and at birth. They have an increased risk of developing diabetes later in life. The babies of women who have been diagnosed with GDM can be larger than normal and this can cause injuries to the mother and the baby at birth. The birth is more likely to be induced or the baby born by caesarean section. These babies are at risk of developing diabetes as children or young adults. Finding the best medications to treat the women and prevent the complications that are linked to GDM is therefore important.

What evidence did we find?

We searched for studies on 14 May 2016. We included 11 randomised controlled trials involving 1487 mothers and their babies (but only eight trials contributed data to our analyses). The evidence was limited by the quality and number of studies and we advise caution when looking at the results.

The criteria for diagnosis of GDM and treatment targets varied between studies, and each outcome is based on few studies with low numbers of women. Three studies compared oral medication with placebo/standard care but the following findings are from a single study (375 women). The quality of the evidence was very low or low. We found no differences between the oral medication and placebo group for the risk of high blood pressure, birth by caesarean section, induction of labour or perineal trauma. The number of babies born large‐for‐gestational age, with low blood sugars or dying at birth was not clearly different between the groups. Two studies (434 women) reported no difference in the need for insulin between the oral medication and placebo group.

Six studies compared metformin with glibenclamide. The quality of the evidence was very low to moderate. We found no difference between metformin and glibenclamide for the risk of high blood pressure (three studies, 508 women, moderate‐quality evidence), birth by caesarean section (four studies, 554 women, low‐quality evidence), perineal trauma (two studies, 308 women, low‐quality evidence) or induction of labour (one study, 159 women, low‐quality evidence). We found no difference between metformin and glibenclamide for the baby having low blood sugars (four studies, 554 infants, low‐quality evidence), being born large‐for‐gestational age (two studies, 246 infants) or dying at birth (all low‐ or very low‐quality evidence). In one study, the babies of the mothers taking metformin were at reduced risk of having any serious outcome (low blood sugar, jaundice, being born large, breathing problems, injury at birth or death combined) (low‐quality evidence). One small study (43 women) comparing glibenclamide with acarbose reported no differences in outcomes for mothers or their babies.

None of the included studies provided any data on many of the outcomes pre‐specified in this review, including long‐term outcomes for the mother or for the baby as a child or an adult.

What does this mean?

There is not enough high‐quality evidence available to guide us on if oral medication has better outcomes for women with gestational diabetes, and their babies, compared with a placebo or if one oral medication has better health outcomes than another oral medication. Because we are still unclear, further research is needed. Future studies should be encouraged to report on the outcomes suggested in this review and in particular the long‐term outcomes for the woman and the infant that have been poorly reported to date.

Summary of findings

Summary of findings for the main comparison. Oral anti‐diabetic pharmacological therapies versus placebo ‐ maternal outcomes.

Oral anti‐diabetic pharmacological therapies versus placebo ‐ maternal outcomes
Patient or population: women diagnosed with gestational diabetes; 24‐30 weeks' gestation; singleton pregnancy. Setting: Medical Centre, USA. Intervention: oral anti‐diabetic pharmacological therapy (glibenclamide) Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with placebo	Risk with oral anti‐diabetic pharmacological therapies
Hypertensive disorders of pregnancy ‐ (any type)	167 per 1000	207 per 1000 (135 to 317)	RR 1.24 (0.81 to 1.90)	375 (1 RCT)	⊕⊝⊝⊝ Very lowabc
Caearean section	360 per 1000	371 per 1000 (285 to 483)	RR 1.03 (0.79 to 1.34)	375 (1 RCT)	⊕⊝⊝⊝ Very lowabc
Development of type 2 diabetes ‐ not measured	see comment	see comment	not estimable	‐	‐	This was not a pre‐specified outcome for the included studies reporting on this comparison
Perineal trauma	5 per 1000	5 per 1000 (0 to 84)	RR 0.98 (0.06 to 15.62)	375 (1 RCT)	⊕⊝⊝⊝ Very lowabcd	Event rates were low 1/189 for anti‐diabetic pharmacological therapy and 1/186 in the control (placebo) group
Return to pre‐pregnancy weight ‐ not measured	see comment	see comment	not estimable	‐	‐	This was not a pre‐specified outcome for the included studies reporting on this comparison
Postnatal depression ‐ not measured	see comment	see comment	not estimable	‐	‐	This was not a pre‐specified outcome for the included studies reporting on this comparison
Induction of labour	188 per 1000	222 per 1000 (149 to 331)	RR 1.18 (0.79 to 1.76)	375 (1 RCT)	⊕⊝⊝⊝ Very lowabc
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio; OR: odds ratio
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

^aRisk of bias ‐ we did not find a published protocol and there were more outcomes reported in the published paper than were listed in the trial registration document ‐ downgraded 1 level.
 ^bGeneralisability ‐ in this single study 93% of participants were Hispanic women. Results may not be generalisable to other populations ‐ downgraded 1 level.
 ^cImprecision ‐ evidence based on a single study ‐ downgraded 1 level.
 ^dImprecision ‐ wide confidence intervals crossing the line of no effect and low event rates suggestive of imprecision ‐ downgraded 1 level.

Summary of findings 2. Oral anti‐diabetic pharmacological therapies versus placebo ‐ neonatal outcomes.

Oral anti‐diabetic pharmacological therapies versus placebo ‐ neonatal outcomes
Patient or population: infants of women diagnosed with gestational diabetes Setting: Medical Centre, USA Intervention: oral anti‐diabetic pharmacological therapies (glibenclamide) Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with placebo	Risk with oral anti‐diabetic pharmacological therapies
Large‐for‐gestational age	118 per 1000	105 per 1000 (60 to 187)	RR 0.89 (0.51 to 1.58)	375 (1 RCT)	⊕⊝⊝⊝ Very lowabc
Perinatal mortality	see comment	see comment	not estimable	‐	‐	None of the included studies pre‐specified this outcome
Death or serious morbidity composite ‐ not reported	see comment	see comment	not estimable	‐	‐	None of the included studies pre‐specified this outcome
Neonatal hypoglycaemia	11 per 1000	21 per 1000 (4 to 114)	RR 1.97 (0.36 to 10.62)	375 (1 RCT)	⊕⊝⊝⊝ Very lowabcd	Event rates were low with 4/189 for oral anti‐diabetic pharmacological therapy and 2/186 for placebo group
Adiposity (neonate, child, adult) ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included studies pre‐specified this outcome
Diabetes (child, adult) ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included studies pre‐specified this outcome
Neurosensory disability in later childhood ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included studies pre‐specified this outcome
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio; OR: odds ratio
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

^aRisk of bias ‐ we did not find a published protocol and there were more outcomes reported in the published paper than were listed in the trial registration document ‐ downgraded 1 level.
 ^bThis single trial comprised 93% Hispanic women. The results may not be generalisable ‐ downgraded 1 level.
 ^cImprecision ‐ evidence was based on a single trial only ‐ downgraded 1 level.
 ^dImprecison ‐ wide confidence intervals crossing the line of no effect and low event rates ‐ downgraded 1 level.

Summary of findings 3. Metformin versus glibenclamide ‐ maternal outcomes.

Metformin versus glibenclamide ‐ maternal outcomes
Patient or population: women diagnosed with gestational diabetes Setting: trials conducted in Brazil, India and the USA Intervention: metformin Comparison: glibenclamide
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with glibenclamide	Risk with metformin
Hypertensive disorders of pregnancy	88 per 1000	62 per 1000 (33 to 114)	RR 0.70 (0.38 to 1.30)	508 (3 RCTs)	⊕⊕⊕⊝ Moderatea
Caesarean section	392 per 1000	470 per 1000 (325 to 674)	RR 1.20 (0.83 to 1.72)	554 (4 RCTs)	⊕⊕⊝⊝ Lowbc
Development of type 2 diabetes ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included studies for this comparison had pre‐specified development of type 2 diabetes as an outcome
Perineal trauma	6 per 1000	11 per 1000 (1 to 81)	RR 1.67 (0.22 to 12.52)	308 (2 RCTs)	⊕⊕⊝⊝ Lowad	Note low event rates (2/154 for metformin and 1/154 for glibenclamide
Return to pre‐pregnancy weight ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included studies for this comparison had return to pre‐pregnancy weight as a pre‐specified outcome
Postnatal depression ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included studies for this comparison had postnatal depression as a pre‐specified outcome
Induction of labour	613 per 1000	496 per 1000 (374 to 655)	RR 0.81 (0.61 to 1.07)	159 (1 RCT)	⊕⊕⊝⊝ Lowae
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio; OR: odds ratio
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

^aRisk of bias ‐ all studies were open label ‐ downgraded 1 level.
 ^bRisk of bias ‐ 3 of the 4 studies were open label and 3 of 4 studies were unclear for blinding of outcome assessors. 2 studies reported additional outcomes that were not pre‐specified ‐ downgraded 1 level.
 ^cInconsistency ‐ heterogeneity was high, I² = 61% downgraded 1 level.
 ^dImprecision ‐ wide confidence intervals along with low event rates suggest imprecision ‐ downgraded 1 level.
 ^eImprecision ‐ evidence was based on a single trial ‐ downgraded 1 level.

Summary of findings 4. Metformin versus glibenclamide ‐ neonatal outcomes.

Metformin versus glibenclamide ‐ neonatal outcomes
Patient or population: Infants of women diagnosed with gestational diabetes Setting: trials conducted in Brazil, India and the USA Intervention: metformin Comparison: glibenclamide
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with glibenclamide	Risk with metformin
Large‐for‐gestational age	193 per 1000	129 per 1000 (46 to 354)	RR 0.67 (0.24 to 1.83)	246 (2 RCTs)	⊕⊕⊝⊝ Lowab
Perinatal mortality	6 per 1000	5 per 1000 (0 to 83)	RR 0.92 (0.06 to 14.55)	359 (2 RCTs)	⊕⊝⊝⊝ Very lowc	Note that event rates were very low. 1 study had no event of perinatal death in either the metformin nor the glibenclamide group. The second study had 1 death in each group
Death or serious morbidity composite	350 per 1000	189 per 1000 (109 to 329)	RR 0.54 (0.31 to 0.94)	159 (1 RCT)	⊕⊕⊝⊝ Lowd
Neonatal hypoglycaemia	48 per 1000	41 per 1000 (20 to 84)	RR 0.86 (0.42 to 1.77)	554 (4 RCTs)	⊕⊕⊝⊝ Lowae
Adiposity ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included trials for this comparison had pre‐specified adiposity as a trial outcome
Diabetes ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included trials for this comparison had pre‐specified diabetes as a trial outcome
Neurosensory disability in later childhood ‐ not measured	see comment	see comment	not estimable	‐	‐	None of the included trials for this comparison had pre‐specified neurosensory disability as a trial outcome
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio; OR: odds ratio
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

^aRisk of bias ‐ allocation concealment was unclear in 1 study. 1 study was open label ‐ downgraded 1 level.
 ^bInconsistency ‐ heterogeneity was I²= 54%, which could not be explained by the diagnostic criteria used ‐ downgraded 1 level.
 ^cRisk of bias ‐ includes open label study/studies with no evidence of blinding of participants or researchers ‐ downgraded 1 level.
 ^dImprecision ‐ evidence based on a single small study ‐ downgraded 1 level.
 ^eImprecision ‐ event rates low (< 30) ‐ downgraded 1 level.

Summary of findings 5. Glibenclamide versus acarbose ‐ maternal outcomes.

Glibenclamide versus acarbose ‐ maternal outcomes
Patient or population: women diagnosed with gestational diabetes; 11 to 33 weeks' gestation; singleton pregnancy Setting: Maternity hospital, Brazil Intervention: Glibenclamide Comparison: Acarbose
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with acarbose	Risk with other oral anti‐diabetic agent
Hypertensive disorders of pregnancy ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
Caesarean section	526 per 1000	500 per 1000 (279 to 895)	RR 0.95 (0.53 to 1.70)	43 (1 RCT)	⊕⊕⊝⊝ Lowab
Development of type 2 diabetes ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
Perineal trauma ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
Return to pre‐pregnancy weight ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
Postnatal depression ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
Induction of labour ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio; OR: odds ratio
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

^aMethod of randomisation was unclear and the study was open‐label ‐ downgraded ‐1 level.
 ^bEvidence based on a single small study ‐ downgraded ‐1 level.

Summary of findings 6. Glibenclamide versus acarbose ‐ neonatal outcomes.

Glibenclamide versus acarbose ‐ neonatal outcomes
Patient or population: women with gestational diabetes Setting: maternity hospital, Brazil Intervention: glibenclamide Comparison: acarbose
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with acarbose	Risk with glibenclamide
Large‐for‐gestational age	105 per 1000	251 per 1000 (57 to 1000)	RR 2.38 (0.54 to 10.46)	43 (1 RCT)	⊕⊕⊝⊝ Lowab
Perinatal mortality	see comment	see comment	not estimable	43 (1 RCT)	⊕⊕⊝⊝ Lowab	No events were reported in either group
Death or serious morbidity composite ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
Neonatal hypoglycaemia	53/1000	333/1000 (46 to 1000)	RR 6.33 (0.87 to 46.32)	43 (1 RCT)	⊕⊕⊝⊝ Lowab	Low event rates and sample size (8/24 in glibenclamide group and 1/19 in acarbose group)
Adiposity ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
Diabetes ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
Neurosensory disability in later childhood ‐ not reported	see comment	see comment	not estimable	‐	‐	No data were reported for this outcome
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio; OR: odds ratio
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

^aRisk of bias ‐ evidence of selective reporting ‐ downgraded 1 level.
 ^bImprecision ‐ evidence based on a single small study with wide confidence intervals ‐ downgraded 1 level.

Background

The original review by Alwan and colleagues, Treatments for gestational diabetes (Alwan 2009) has been split into three new review titles reflecting the complexity of treating women with gestational diabetes.

• Lifestyle interventions for the treatment of women with gestational diabetes mellitus (Brown 2015)

• Oral anti‐diabetic pharmacological therapies for the treatment of women with gestational diabetes mellitus (this review)

• Insulin for the treatment of women with gestational diabetes mellitus (Brown 2016)

There will be similarities in the background, methods and outcomes between these three systematic reviews. Portions of the methods section of this review are based on a standard template used by Cochrane Pregnancy and Childbirth.

Description of the condition

Gestational diabetes mellitus (GDM) often referred to as gestational diabetes can be defined as "glucose intolerance or hyperglycaemia (high blood glucose concentration) with onset or first recognition during pregnancy" (WHO 1999). GDM occurs when the body is unable to make enough insulin to meet the extra needs in pregnancy. The high blood sugars associated with GDM will return to normal after the birth of the baby. However, there are currently no universally accepted diagnostic criteria (ACOG 2013; Coustan 2010; HAPO 2008; Hoffman 1998; IADPSG 2010; Metzger 1998; NICE 2015). GDM may include previously undetected type 1 diabetes, type 2 diabetes or diabetes presenting only during pregnancy (HAPO 2008; IADPSG 2010; Metzger 1998; Nankervis 2014; WHO 2014).

GDM is one of the most common pregnancy complications and the prevalence is rising worldwide with 1% to 36% of pregnancies being affected (Bottalico 2007; Cundy 2014; Duran 2014; Ferrara 2007; NICE 2015; Ragnarsdottir 2010; Tran 2013). The prevalence of GDM is likely to continue to increase along with the increasing prevalence of maternal obesity and associated type 2 diabetes (Bottalico 2007; Mulla 2010; Petry 2010).

A variety of factors have been associated with an increased risk of developing GDM. Non‐modifiable risk factors include advanced maternal age (Chamberlain 2013; Morisset 2010), high parity, non‐white race or ethnicity (in particular South Asian, Middle Eastern), family history of diabetes, maternal high or low birthweight, polycystic ovarian syndrome (Cypryk 2008; Petry 2010; Solomon 1997), a history of having a previous macrosomic infant (birthweight 4000 g or more) and previous history of GDM (Petry 2010).

Modifiable risk factors include physical inactivity (Chasan‐Taber 2008), having a low‐fibre and high‐glycaemic load diet (Zhang 2006), maternal overweight (body mass index (BMI) equal to or greater than 25 kg/m²) or obesity (equal to or greater than 30 kg/m²) (Kim 2010) and excessive weight gain during pregnancy, especially for those who are already overweight or obese (Hedderson 2010).

Pathophysiology of gestational diabetes

Normal pregnancy is associated with significant changes in maternal metabolism (Lain 2007). In early pregnancy, oestrogen and progesterone stimulate maternal beta‐cell hyperplasia and insulin secretion, which promotes maternal nutrient storage (adipose and hepatic glycogen) to support later fetal growth. At this stage, insulin sensitivity is maintained or may even increase. However, as pregnancy progresses whole‐body insulin sensitivity steadily decreases, such that by the third trimester it is reduced by almost half (Barbour 2007). Several factors contribute to this, including placental hormones (human placental lactogen and placental growth hormone), cytokines released from adipocytes (IL‐6, TNF‐alpha), increased free fatty acids and lower adiponectin concentrations (Clapp 2006; Devlieger 2008). This results in decreased post‐prandial peripheral glucose disposal by up to 40% to 60% (Barbour 2007). Because glucose is transported to the fetus by facilitated diffusion, this state of physiological insulin resistance promotes fetal glucose uptake, a principal oxidative fuel and carbon source for the growing fetus. In normal pregnancy, maternal glycaemia is maintained by a significant increase in insulin secretion of up to 200% to 250% (Barbour 2007; Lain 2007; Suman Rao 2013).

Regulation of fetal glucose metabolism requires (1) the maintenance of maternal glucose concentration through increasing maternal glucose production, and at the same time, developing maternal glucose intolerance and insulin resistance, (2) transfer of glucose to the fetus across the placenta and (3) production of fetal insulin and uptake of glucose into adipose tissue and skeletal muscle (Suman Rao 2013).

Women with GDM have further reductions in insulin signalling, and glucose uptake is decreased beyond that of normal pregnancy (Barbour 2007). This results in glucose intolerance, though glycaemia in pregnancy represents a continuum. In GDM, the steeper maternal‐fetal glucose gradient, especially post‐prandial, leads to increased fetal glucose uptake which stimulates fetal insulin secretion. Insulin is a key fetal anabolic hormone and hyperinsulinaemia promotes fetal overgrowth leading to large‐for‐gestational age (LGA) infants, macrosomia, and possible organ damage (Catalano 2003; Ju 2008; Metzger 2008; Reece 2009).

Women with GDM also have increased circulating inflammatory cytokines and lower adiponectin concentrations leading to increased lipolysis and fatty acid concentrations. Placental transfer of free fatty acids contributes to increased fetal adiposity, independent of glucose uptake (Knopp 1985). Thus, even women with well‐controlled GDM still have increased risk of fetal macrosomia (Langer 2005).

Screening and diagnosis of GDM

Regardless of whether universal or selective (risk‐factor), screening with a 50 g oral glucose challenge test is used, diagnosis of GDM is usually based on either a 75 g two‐hour oral glucose tolerance test (OGTT) or a 100 g three‐hour OGTT (ADA 2013; IADPSG 2010; Nankervis 2014; NICE 2015; WHO 1999). Recommendations regarding diagnostic criteria vary nationally and internationally (Table 7), and these diagnostic criteria have changed over time, sometimes due to changing understanding about the effects of hyperglycaemia on pregnancy and infant outcomes (Coustan 2010), but also because of a lack of evidence clearly demonstrating the clinical and cost‐effectiveness of one criterion over another.

1. Examples of diagnostic criteria for gestational diabetes.

Organisation/professional body	Screening and diagnostic criteria
	1‐hour oral glucose challenge test	Oral glucose tolerance test	Fasting	1 hour	2 hour	3 hour
ADA 2013a, IADPSG 2010a, ADIPS 2013 (Nankervis 2014)a, WHO 2014a	‐	75 g	≥ 5.1 mmol/L (≥ 92 mg/dL)	≥ 10 mmol/L (≥ 180 mg/dL)	≥ 8.5 mmol/L (≥ 153 mg/dL)	‐
ACOG 2013 Carpenter and Coustanb National Diabetes Data Groupb	50 g (> 7.2 mmol/L; > 130 mg/dL)	100 g	≥ 5.3 mmol/L (≥ 95 mg/dL)	≥ 10 mmol/L (≥180 mg/dL)	≥ 8.6 mmol/L (≥ 155 mg/dL)	≥ 7.8 mmol/L (≥ 140 mg/dL)
	50 g (> 7.8 mmol/L; > 140 mg/dL)	100 g	≥ 5.8 mmol/L (≥ 105 mg/dL)	≥ 10.6 mmol/L (≥ 190 mg/dL)	≥ 9.2 mmol/L (≥ 165 mg/dL)	≥ 8.0 mmol/L (≥ 145 mg/dL)
Canadian Diabetes Association 2013 eithera orb	50 g ‐	75 g 75 g	≥ 5.3 mmol/L (≥ 95 mg/dL) ≥ 5.1 mmol/L (≥ 92 mg/dL)	≥ 10.6 mmol/L (≥ 190 mg/dL) ≥ 10 mmol/L (≥ 180 mg/dL)	≥ 9.0 mmol/L ≥ 8.5 mmol/L (≥ 153 mg/dL)
NICE 2015	‐	75 g	≥ 5.6 mmol/L (≥ 101 mg/dL)	‐	≥ 7.8 mmol/L (≥ 140 mg/dL)	‐
NICE 2008; WHO 1999; Hoffman 1998 (ADIPS)b	‐	75 g	≥ 7.0 mmol/L (≥ 126 mg/dL)	‐	≥ 11.1 mmol/L (≥ 200 mg/dL)	‐
New Zealand Ministry of Health 2014a	50 g if HbA1c < 41 mmol/mol (≥ 7.8 mmol/L; ≥ 140 mg/dL)	75 g	≥ 5.5 mmol/L (≥ 99 mg/dL)	‐	≥ 9.0 mmol/L (≥ 162 mg/dL)	‐

ADA: American Diabetes Association
 IADPSG: International Association of the Diabetes and Pregnancy Study Groups
 ADIPS: Australasian Diabetes in Pregnancy Society
 ACOG: American College of Obstetrics and Gynecology
 NICE: National Institute for Health and Care Excellence
 ^a1 abnormal result required for diagnosis
 ^b2 or more abnormal results required for diagnosis

The Hyperglycaemia and Adverse Pregnancy Outcomes (HAPO) study (HAPO 2008), a large, international observational study reported graded linear associations in the odds of several GDM‐associated adverse outcomes and glucose levels at OGTT, with no clear threshold identified at which risk increased substantially. The International Association of the Diabetes and Pregnancy Study Groups (IADPSG) recommended diagnostic criteria using data from the HAPO study (IADPSG 2010). Applying the IADPSG criteria in most health environments will increase the number of women with GDM. A study conducted in Vietnam showed that depending on the criteria used, the diagnosis of GDM varied between 5.9% (American Diabetes Association ‐ ADA), 20.4% (International Association of Diabetes in Pregnancy Study Groups ‐ IADPSG), 20.8% (Australasian Diabetes in Pregnancy Society ‐ ADIPS), and up to 24.3% (World Health Organization ‐ WHO) (Tran 2013). A Bulgarian study also reported differences in prevalence based on the diagnostic criteria ranging from 10.8% (European Association for the Study of Diabetes ‐ EASD), 13.5% (ADA), 16.2% (New Zealand Society for the Study of Diabetes ‐ NZSSD), 17.1% (WHO), 21.2% (ADIPS), 31.6% (IADPSG) (Boyadzhieva 2012).

Clinical outcomes for women with gestational diabetes

Adverse outcomes have been consistently reported at higher rates in women diagnosed with GDM and their infants compared to women without GDM (Crowther 2005; Landon 2009; Metzger 2008; Reece 2009).

Women with GDM have an increased risk of developing pre‐eclampsia, are more likely to have their labour induced (Anderberg 2010; Crowther 2005; Ju 2008; Landon 2009; Metzger 2008), and give birth by caesarean section (Landon 2009; Metzger 2008). The incidence of uterine rupture, shoulder dystocia and perineal lacerations are increased in women with GDM due to the increased likelihood of having a LGA or macrosomic baby (Jastrow 2010). Women who have experienced GDM are at a greater risk of metabolic dysfunction in later life (Shah 2008; Vohr 2008), with a crude cumulative incidence of type 2 diabetes of 10% to 20% within 10 years (Bellamy 2009; Kim 2002), but up to 50% when adjusted for retention and length of follow‐up (Kim 2002).

Neonatal, infant and later outcomes related to gestational diabetes

A significant adverse health outcome for babies born to mothers with GDM is being born LGA or macrosomic (Catalano 2003; Crowther 2005; Landon 2009; Metzger 2008; Reece 2009). Large‐for‐gestational age or macrosomic infants are at increased risk of birth injury, such as shoulder dystocia, perinatal asphyxia, bone fractures and nerve palsies (Esakoff 2009; Henriksen 2008; Langer 2005; Metzger 2008).

Babies born to women with GDM, compared with babies born to women without GDM, have significantly greater skinfold measures and fat mass compared with infants of women with normal glucose tolerance (Catalano 2003). The offspring of women with GDM are heavier (adjusted for height) and have greater adiposity than the offspring of women with normal glycaemia during pregnancy (Pettitt 1985; Pettitt 1993), and are more likely to develop early overweight or obesity, type 2 diabetes (Hillier 2007; Pettitt 1993; Whincup 2008), or metabolic syndrome (a cluster of risk factors defined by the occurrence of three of the following: obesity, hypertension, hypertriglyceridaemia and low concentration of high‐density lipoprotein (HDL) cholesterol) in childhood, adolescence or adulthood (Guerrero‐Romero 2010; Harder 2009).

The development of the metabolic syndrome during childhood is a risk factor for the development of adult type 2 diabetes at 25 to 30 years of age (Morrison 2008). These health problems repeat across generations (Dabelea 2005; Mulla 2010) and are important from a public health perspective, because with each generation the prevalence of diabetes increases. Other adverse outcomes which are increased for babies born to women with GDM include respiratory distress syndrome, hypoglycaemia (which if prolonged can cause brain injury), hyperbilirubinaemia, hypertrophic cardiomyopathy, hypocalcaemia, hypomagnesaemia, polycythaemia and admission to the neonatal nursery (Metzger 2008; Reece 2009).

Description of the intervention

First‐line treatment for women with GDM is usually a lifestyle intervention combining dietary and exercise components. Where glycaemic treatment targets are not attained by lifestyle interventions alone, or where initial glucose levels are considered to be very high, the option for treatment is to introduce a pharmacological intervention. Current options include subcutaneous insulin or oral anti‐diabetic pharmacological therapies.

There has been an increase in the use of oral anti‐diabetic pharmacological therapies as an alternative to subcutaneous insulin (Ogunyemi 2011) for the treatment of women with GDM, due to lower costs, ease of administration and acceptability (Ryu 2014). The most commonly used therapies are glyburide (glibenclamide) and metformin, although there are other less frequently used anti‐diabetic pharmacological therapies such as acarbose (Kalra 2015). Despite their wide use, oral antidiabetic pharmacological therapies are not licensed for use during pregnancy in many countries (including Australia, New Zealand, UK, USA).

First generation sulphonylureas

First generation sulphonylureas, such as chlorpropamide and tolbutamide have been used to treat diabetes in pregnancy in the past. Both drugs cross the placenta and have been associated with prolonged neonatal hypoglycaemia (Christesen 1998; Kemball 1970).

Second generation sulphonylureas

Second generation sulphonylureas, such as glibenclamide (glyburide) or glipizide work by enhancing insulin secretion. In pregnancy, studies have focused on the use of glibenclamide as it was shown to have the least placental passage in vitro (Elliott 1991). However, with improved assays it is now estimated that cord plasma concentrations of glibenclamide can be 70% to 77% of maternal levels (Schwarz 2013). Glibenclamide is completely metabolised by the liver and the metabolites are excreted equally in bile and urine. Oral tablets are administered at a typical dose of 2.5 to 5 mg taken once a day. Dosage can be increased if glycaemic control is not achieved in increments of 2.5 mg daily at intervals of every seven days (Brayfield 2014). The maximum daily dose is usually 15 mg taken in divided doses. Glibenclamide is associated with weight gain during pregnancy and postpartum and maternal hypoglycaemia (Simmons 2015). Other commonly reported side effects include nausea, vomiting, sensations of fullness, abdominal pain, anorexia, heartburn and diarrhoea. Less common side effects include abnormal liver function, haematological reactions and dermatological reactions (www.medsafe.govt.nz). Glibenclamide is contraindicated in cases of renal and hepatic insufficiency (www.medsafe.govt.nz). Langer 2000 reported equivalence between glyburide and subcutaneous insulin for the primary outcome of glycaemic control and reported that glibenclamide was not detected in the cord blood of any of the babies whose mother had been treated with glibenclamide in their trial.

Metformin

Metformin is a biguanide that can be administered in immediate‐release or sustained‐release oral preparations. The drug is absorbed along the entire gastrointestinal mucosa, but absorption is incomplete. Typical doses and dose scheduling of metformin provide steady‐state plasma concentrations within 24 to 48 hours of administration and are generally less than 1 microgram/mL. Metformin is excreted unchanged in the urine and is not metabolised by the liver. Tablets should be taken in divided doses with meals. The initial dose is typically 500 mg taken once or twice daily and may be increased over subsequent weeks, dependent on glycaemic control, up to a maximum of 1000 mg three times per day (www.medsafe.govt.nz). In clinical trials using metformin during pregnancy the maximum dose is 2500 mg daily (Rowan 2008). Renal insufficiency is reported as a contra‐indication to the prescription of metformin (www.fda.gov).

Metformin is known to cross the placenta although there is no evidence to suggest that this leads to fetal abnormalities (Ekpebegh 2007; Gilbert 2006). Maternal lactic acidosis is a rare (0.03 cases per 1000 patient years) but is a serious metabolic condition that is associated with accumulation of metformin during treatment (www.medsafe.govt.nz). Metformin is commonly associated with gastrointestinal disturbances (diarrhoea, nausea, vomiting) (Simmons 2015). Mild erythema, reduced vitamin B12 absorption and a metallic taste are also reported as side effects of metformin (www.medsafe.govt.nz). As very little metformin is transferred to human breast milk it is thought to be safe during breastfeeding (Simmons 2015). A sentinel trial conducted by Rowan 2008 reported equivalence between metformin and subcutaneous insulin for maternal and infant outcomes.

Combined metformin and glibenclamide

Metformin and glibenclamide are available in some regions in a combined formulation (Europe, USA).

Acarbose

Acarbose is an alpha‐glucosidase inhibitor that delays the digestion of carbohydrates, thus resulting in a reduced increase in post‐prandial blood glucose concentrations (www.accessdata.fda.gov/scripts/cder/daf/). The typical initial dose is 75 to 150 mg taken in three divided doses. This can be further increased to a maximum of 600 mg per day (in divided doses) after four to eight weeks. Acarbose is metabolised within the gastrointestinal tract and the small amount that is excreted is via the urine (www.medsafe.govt.nz; www.accessdata.fda.gov/scripts/cder/daf/). Animal studies have not found an association between acarbose and fetal teratogenicity, but data are lacking in human studies (Kalra 2015). Acarbose is not associated with maternal hypoglycaemia but is frequently associated with gastrointestinal side effects (Simmons 2015). Acarbose is contraindicated in people with inflammatory bowel disease or similar conditions, malabsorption syndromes, severe hepatic or renal impairment (www.medsafe.govt.nz).

Oral insulin

Oral insulin is currently under research development for the treatment of diabetes mellitus, which may include GDM (Fonte 2013; Iyer 2010). This systematic review will not include trials of oral insulin as they will be included in the review of Insulin for the treatment of women with gestational diabetes (Brown 2016).

How the intervention might work

Glibenclamide

Glibenclamide stimulates increased insulin secretion by binding to pancreatic beta‐cell receptors. There is reduced basal hepatic glucose production and enhancement of peripheral insulin action at post‐receptor (probably intracellular) sites. The mechanism of action of glibenclamide requires functional beta‐cells (Patanè 2000). Glibenclamide also inhibits glucagon‐producing alpha cells in the pancreas and increases the release of somatostatin. There is a mild diuretic action which increases free water clearance (Radó 1974).

Metformin

Metformin actions are not completely understood but it inhibits glucogenesis in the liver, delays glucose absorption from the gastrointestinal tract and stimulates glucose uptake into the peripheral tissues. It has been suggested that as metformin increases insulin sensitivity, it does not stimulate insulin release, but does require the presence of insulin to potentiate its antihyperglycaemic effect. Both fasting and post‐prandial blood glucose are lowered in individuals with diabetes. As metformin does not stimulate insulin secretion it is not associated with an increase in hypoglycaemia in diabetic or non‐diabetic populations (www.medsafe.govt.nz). Metformin may protect β‐cell function in the baby/child/adult and as a consequence reduce the cross generational effects of obesity and type 2 diabetes (Simmons 2015). During pregnancy the pharmacological action of metformin is altered slightly due to enhanced renal elimination, varying food absorption and gastrointestinal transit times (Simmons 2015).

Acarbose

Acarbose does not enhance insulin secretion. It reduces intestinal carbohydrate absorption by inhibiting the cleavage of disaccharides and oligosaccharides to monosaccharides in the small intestine resulting in reduced glucose absorption and lower post‐prandial glucose concentrations (Kalra 2015).

Why it is important to do this review

Although oral anti‐diabetic pharmacological therapies are more acceptable to women with GDM (Rowan 2008), cost less and are easier to administer than subcutaneous insulin, the safety of these therapies is still unclear.

The comparison of subcutaneous or oral insulin with oral anti‐diabetic pharmacological therapies is the subject of a new Cochrane Review and will not therefore be covered in this review.

The superiority of one anti‐diabetic pharmacological therapy over another has not previously been addressed by a Cochrane systematic review.

Objectives

To evaluate the effects of oral anti‐diabetic pharmacological therapies for treating women with GDM.

Methods

Criteria for considering studies for this review

Types of studies

We included published or unpublished randomised, quasi‐randomised or cluster‐randomised trials in full text or abstract format. We excluded cross‐over trials. Conference abstracts were handled in the same way as full publications.

Types of participants

Participants were pregnant women diagnosed with gestational diabetes mellitus (GDM) (diagnosis as defined by the individual trial). Women with type 1 or type 2 diabetes diagnosed prior to pregnancy were excluded.

Types of interventions

We included those anti‐diabetic pharmacological therapies that were used during pregnancy including metformin, glibenclamide, acarbose, tolbutamide, chlorpropamide or any combination of these therapies. Only oral anti‐diabetic pharmacological therapies were included in this review. We will include any new therapies prescribed for the treatment of women with GDM in updates of the review.

• Oral anti‐diabetic pharmacological therapy versus placebo or no pharmacological treatment.

• A single oral anti‐diabetic pharmacological therapy versus an alternative oral anti‐diabetic pharmacological therapy (drug A versus drug B).

• Any combination of oral anti‐diabetic pharmacological therapies versus any combination of oral anti‐diabetic pharmacological therapies (drug A followed by drug B versus drug B followed by drug A for example).

• Oral anti‐diabetic pharmacological therapy versus another intervention (excluding insulin) not specified above.

We have not included the comparison of oral anti‐diabetic pharmacological therapies versus insulin in this review as it will be covered in the review entitled Insulin for the treatment of women with gestational diabetes (Brown 2016).

Types of outcome measures

Primary outcomes

Maternal

• Hypertensive disorders of pregnancy (including pre‐eclampsia, pregnancy‐induced hypertension, eclampsia as defined by trialists)

• Caesarean section

• Development of type 2 diabetes

Neonatal

• Perinatal (fetal and neonatal death) and later infant mortality

• Large‐for‐gestational age (LGA) (as defined by trialists)

• Death or serious morbidity composite (variously defined by trials, e.g. perinatal or infant death, shoulder dystocia, bone fracture or nerve palsy)

• Neurosensory disability in later childhood (as defined by trialists)

Secondary outcomes

Maternal

• Use of additional pharmacotherapy

• Maternal hypoglycaemia (as defined by trialists)

• Glycaemic control during/end of treatment (as defined by trialists)

• Weight gain in pregnancy

• Adherence to the intervention

• Induction of labour

• Placental abruption

• Postpartum haemorrhage (as defined by trialists)

• Postpartum infection

• Perineal trauma/tearing

• Breastfeeding at discharge, six weeks postpartum, six months or longer

• Maternal mortality

• Sense of well‐being and quality of life

• Behavioural changes associated with the intervention

• Views of the intervention

• Relevant biomarker changes associated with the intervention (including adiponectin, free fatty acids, triglycerides, high‐density lipoproteins, low‐density lipoproteins, insulin)

Long‐term outcomes for mother

• Postnatal depression

• Body mass index (BMI)

• Postnatal weight retention or return to pre‐pregnancy weight

• Type 1 diabetes

• Type 2 diabetes

• Impaired glucose tolerance

• Subsequent gestational diabetes

• Cardiovascular health (as defined by trialists including blood pressure, hypertension, cardiovascular disease, metabolic syndrome)

Fetal/neonatal outcomes

• Stillbirth

• Neonatal death

• Macrosomia (greater than 4000 g; or as defined by individual study)

• Small‐for‐gestational age (as defined by trialists)

• Birth trauma (shoulder dystocia, bone fracture, nerve palsy)

• Gestational age at birth

• Preterm birth (less than 37 weeks’ gestation; and less than 32 weeks' gestation)

• Five‐minute Apgar less than seven

• Birthweight and z score

• Head circumference and z score

• Length and z score

• Ponderal index

• Adiposity (including skinfold thickness measurements (mm), fat mass)

• Neonatal hypoglycaemia (as defined by trialists)

• Respiratory distress syndrome

• Neonatal jaundice (hyperbilirubinaemia) (as defined by trialists)

• Hypocalcaemia (as defined by trialists)

• Polycythaemia (as defined by trialists)

• Relevant biomarker changes associated with the intervention (including insulin, cord c‐peptide)

Later Infant/childhood outcomes

• Weight and z scores

• Height and z scores

• Head circumference and z scores

• Adiposity (including BMI, skinfold thickness, fat mass)

• Educational attainment

• Blood pressure

• Type 1 diabetes

• Type 2 diabetes

• Impaired glucose tolerance

• Dyslipidaemia or metabolic syndrome

Child as an adult outcomes

• Weight

• Height

• Adiposity (including BMI, skinfold thickness, fat mass)

• Cardiovascular health (as defined by trialists including blood pressure, hypertension, cardiovascular disease, metabolic syndrome)

• Employment, education and social status/achievement

• Dyslipidaemia or metabolic syndrome

• Type 1 diabetes

• Type 2 diabetes

• Impaired glucose tolerance

Health service use

• Number of antenatal visits or admissions

• Number of hospital or health professional visits (including midwife, obstetrician, physician, dietician, diabetic nurse)

• Admission to neonatal intensive care unit/nursery

• Length of antenatal stay

• Length of postnatal stay (maternal)

• Length of postnatal stay (baby)

• Cost of maternal care

• Cost of offspring care

• Costs associated with the intervention

• Costs to families associated with the management provided

• Cost of dietary monitoring (e.g. diet journals, dietician, nurse visits, etc)

• Costs to families ‐ change of diet, extra antenatal visits

• Extra use of healthcare services (consultations, blood glucose monitoring, length and number of antenatal visits)

• Women’s view of treatment advice

• Duration of stay in neonatal intensive care unit or special care baby unit

• Duration of maternal and neonatal hospital stay (antenatal, neonatal, postnatal)

Search methods for identification of studies

The following methods section of this review is based on a standard template used by Cochrane Pregnancy and Childbirth.

Electronic searches

We searched Cochrane Pregnancy and Childbirth’s Trials Register by contacting their Information Specialist (16 May 2016).

The Register is a database containing over 22,000 reports of controlled trials in the field of pregnancy and childbirth. For full search methods used to populate Pregnancy and Childbirth’s Trials Register including the detailed search strategies for CENTRAL, MEDLINE, Embase and CINAHL; the list of handsearched journals and conference proceedings, and the list of journals reviewed via the current awareness service, please follow this link to the editorial information about the Cochrane Pregnancy and Childbirth in the Cochrane Library and select the ‘Specialized Register ’ section from the options on the left side of the screen.

Briefly, Cochrane Pregnancy and Childbirth’s Trials Register is maintained by their Information Specialist and contains trials identified from:

1. monthly searches of the Cochrane Central Register of Controlled Trials (CENTRAL);

2. weekly searches of MEDLINE (Ovid);

3. weekly searches of Embase (Ovid);

4. monthly searches of CINAHL (EBSCO);

5. handsearches of 30 journals and the proceedings of major conferences;

6. weekly current awareness alerts for a further 44 journals plus monthly BioMed Central email alerts.

Search results are screened by two people and the full text of all relevant trial reports identified through the searching activities described above is reviewed. Based on the intervention described, each trial report is assigned a number that corresponds to a specific Pregnancy and Childbirth review topic (or topics), and is then added to the Register. The Information Specialist searches the Register for each review using this topic number rather than keywords. This results in a more specific search set that has been fully accounted for in the relevant review sections (Included studies; Excluded studies; Studies awaiting classification; Ongoing studies).

In addition, we searched ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (ICTRP) for unpublished, planned and ongoing trial reports (14 May 2016). The search terms we used are given in Appendix 1.

Searching other resources

We searched the reference lists of retrieved studies. We did not apply any language or date restrictions.

Data collection and analysis

The following methods section of this review is based on a standard template used by Cochrane Pregnancy and Childbirth.

Selection of studies

Two review authors independently assessed for inclusion all the potential studies we identified as a result of the search strategy. We resolved any disagreements through discussion. A third person was not required.

We created a study flow diagram to map out the number of records identified, included and excluded (Figure 1).

1.

Study flow diagram

Data extraction and management

We designed a form to extract data. For eligible studies, two review authors extracted the data using the agreed form. We resolved discrepancies through discussion. Consultation with a third person was not required. We entered data into Review Manager (RevMan) software (RevMan 2014) and checked for accuracy. When information regarding any of the above was unclear, we attempted to contact authors of the original reports to provide further details.

Assessment of risk of bias in included studies

Two review authors independently assessed risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). We resolved any disagreement by discussion. Consultation with a third person was not required.

(1) Random sequence generation (checking for possible selection bias)

We described for each included study the method used to generate the allocation sequence in sufficient detail to allow an assessment of whether it should produce comparable groups.

We assessed the method as:

• low risk of bias (any truly random process, e.g. random number table; computer random number generator);

• high risk of bias (any non‐random process, e.g. odd or even date of birth; hospital or clinic record number);

• unclear risk of bias.

(2) Allocation concealment (checking for possible selection bias)

We described for each included study the method used to conceal allocation to interventions prior to assignment and will assess whether intervention allocation could have been foreseen in advance of, or during recruitment, or changed after assignment.

We assessed the methods as:

• low risk of bias (e.g. telephone or central randomisation; consecutively numbered sealed opaque envelopes);

• high risk of bias (open random allocation; unsealed or non‐opaque envelopes, alternation; date of birth);

• unclear risk of bias.

(3.1) Blinding of participants and personnel (checking for possible performance bias)

We described for each included study the methods used, if any, to blind study participants and personnel from knowledge of which intervention a participant received. We considered that studies were at low risk of bias if they were blinded, or if we judge that the lack of blinding would be unlikely to affect results. We assessed blinding separately for different outcomes or classes of outcomes.

We assessed the methods as:

• low, high or unclear risk of bias for participants;

• low, high or unclear risk of bias for personnel.

(3.2) Blinding of outcome assessment (checking for possible detection bias)

We described for each included study the methods used, if any, to blind outcome assessors from knowledge of which intervention a participant received. We assessed blinding separately for different outcomes or classes of outcomes.

We assessed methods used to blind outcome assessment as:

• low, high or unclear risk of bias.

(4) Incomplete outcome data (checking for possible attrition bias due to the amount, nature and handling of incomplete outcome data)

We described for each included study, and for each outcome or class of outcomes, the completeness of data including attrition and exclusions from the analysis. We stated whether attrition and exclusions were reported and the numbers included in the analysis at each stage (compared with the total randomised participants), reasons for attrition or exclusion where reported, and whether missing data were balanced across groups or were related to outcomes. No additional data that could be included in a meta‐analysis were supplied by study authors. In future updates, where sufficient information is reported, or can be supplied by the study authors, we will re‐include missing data in the analyses which we undertake.

We assessed methods as:

• low risk of bias (e.g. no missing outcome data; missing outcome data balanced across groups);

• high risk of bias (e.g. numbers or reasons for missing data imbalanced across groups; ‘as treated’ analysis done with substantial departure of intervention received from that assigned at randomisation);

• unclear risk of bias.

(5) Selective reporting (checking for reporting bias)

We described for each included study how we investigated the possibility of selective outcome reporting bias and what we found.

We assessed the methods as:

• low risk of bias (where it was clear that all of the study’s pre‐specified outcomes and all expected outcomes of interest to the review were reported);

• high risk of bias (where not all the study’s pre‐specified outcomes were reported; one or more reported primary outcomes were not pre‐specified; outcomes of interest were reported incompletely and so could not be used; study failed to include results of a key outcome that would have been expected to have been reported);

• unclear risk of bias.

(6) Other bias (checking for bias due to problems not covered by (1) to (5) above)

We described for each included study any important concerns we had about other possible sources of bias.

We assessed whether each study was free of other problems that could put it at risk of bias:

• low risk of other bias;

• high risk of other bias;

• unclear whether there is risk of other bias.

(7) Overall risk of bias

We made explicit judgements about whether studies were at high risk of bias, according to the criteria given in the Cochrane Handbook for Systematic Reviews of Interventions(Higgins 2011a). With reference to (1) to (6) above, we assessed the likely magnitude and direction of the bias and whether we considered it was likely to impact on the findings. We planned to explore the impact of the level of bias through undertaking sensitivity analyses ‐ see Sensitivity analysis.

Assessment of the quality of the evidence using the GRADE approach

We assessed the quality of the evidence using the GRADE approach as outlined in the GRADE Handbook in order to assess the quality of the body of evidence relating to the following outcomes. We selected up to a maximum of seven outcomes for the mother and seven for the infant covering both short‐ and long‐term outcomes for the main comparisons.

Maternal outcomes

• Hypertensive disorders of pregnancy (including pre‐eclampsia, pregnancy‐induced hypertension, eclampsia)

• Caesarean section

• Development of type 2 diabetes

• Perineal trauma

• Return to pre‐pregnancy weight

• Postnatal depression

• Induction of labour

Neonatal outcomes

• LGA

• Perinatal mortality

• Death or serious morbidity composite (variously defined by studies, e.g. infant death, shoulder dystocia, bone fracture or nerve palsy)

• Neonatal hypoglycaemia

• Adiposity

• Diabetes

• Neurosensory disability in later childhood

We used the GRADEpro Guideline Development Tool (GRADEpro GDT) to import data from RevMan (RevMan 2014) in order to create 'Summary of findings' tables. We produced a summary of the intervention effect and a measure of quality for each of the above outcomes using the GRADE approach. The GRADE approach uses five considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of the body of evidence for each outcome. The evidence can be downgraded from 'high quality' by one level for serious (or by two levels for very serious) limitations, depending on assessments for risk of bias, indirectness of evidence, serious inconsistency, imprecision of effect estimates or potential publication bias.

Measures of treatment effect

Dichotomous data

For dichotomous data, we presented results as summary risk ratio (RR) with 95% confidence intervals (CI).

Continuous data

For continuous data, we used the mean difference (MD) if outcomes were measured in the same way between trials. We used the standardised mean difference (SMD) to combine data from trials that measured the same outcome, but used different methods.

Unit of analysis issues

Cluster‐randomised trials

We did not identify any cluster‐randomised controlled trials in this version of the systematic review. In future updates, if cluster‐randomised trials are identified, we will include cluster‐randomised trials in the analyses along with individually‐randomised trials. We will make adjustments using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Section 16.3.4 or 16.3.6) using an estimate of the intra‐cluster correlation co‐efficient (ICC) derived from the trial (if possible), from a similar trial or from a study of a similar population (Higgins 2011b). If we use ICCs from other sources, we will report this and conduct sensitivity analyses to investigate the effect of variation in the ICC. We will consider it reasonable to combine the results from both cluster‐randomised trials and individually‐randomised trials if there is little heterogeneity between the study designs and the interaction between the effect of intervention and the choice of randomisation unit is considered to be unlikely. If cluster‐randomised trials are included, we will seek statistical advice on appropriate analysis to enable inclusion of data in the meta‐analyses.

Other unit of analysis issues

Multiple pregnancy

There may be unit of analysis issues that arise when the women randomised have a multiple pregnancy. We did not include any studies including multiple pregnancies in this version of the systematic review. In future updates, if multiple pregnancies are identified, we will present maternal data as per woman randomised and neonatal data per infant.

Multiple‐arm studies

Where a trial has multiple intervention arms we planned to avoid 'double counting' of participants by combining groups to create a single pair‐wise comparison if possible. Where this was not possible, we planned to split the 'shared' group into two or more groups with smaller sample size and include two or more (reasonably independent) comparisons. We did not identify any studies of multiple comparisons in this version of the systematic review.

Dealing with missing data

For included studies, we noted levels of attrition. We planned to explore the impact of including studies with high levels of missing data (more than 20%) in the overall assessment of treatment effect by using sensitivity analysis. However, one of the included studies reported attrition greater than 20% (De Bacco 2015) and therefore we did not conduct sensitivity analyses based on attrition in this version of the review. In future updates we plan to explore attrition for sensitivity analysis if sufficient evidence is available.

For all outcomes, we carried out analyses, as far as possible, on an intention‐to‐treat basis, that is, we attempted to include all participants randomised to each group in the analyses, and all participants were analysed in the group to which they were allocated, regardless of whether or not they received the allocated intervention. The denominator for each outcome in each trial was the number randomised minus any participants whose outcomes were known to be missing.

Assessment of heterogeneity

We assessed statistical heterogeneity in each meta‐analysis using the Tau², I² (Higgins 2003) and Chi² statistics. We regarded heterogeneity as substantial if an I² was greater than 40% and either a Tau² was greater than zero, or there was a low P value (less than 0.10) in the Chi² test for heterogeneity.

Assessment of reporting biases

If there were 10 or more studies in the meta‐analysis, we planned to investigate reporting biases (such as publication bias) using funnel plots. This was not possible in this version of the review. In future updates, where possible, we will assess funnel plot asymmetry visually. If asymmetry is suggested by a visual assessment, we will perform exploratory analyses to investigate it.

Data synthesis

We carried out statistical analysis using the Review Manager software (RevMan 2014). We used fixed‐effect meta‐analysis for combining data where it was reasonable to assume that studies were estimating the same underlying treatment effect: that is, where trials were examining the same intervention, and the trials’ populations and methods were judged sufficiently similar. If there was clinical heterogeneity sufficient to expect that the underlying treatment effects differed between trials, or if substantial statistical heterogeneity was detected, we used random‐effects meta‐analysis to produce an overall summary, if an average treatment effect across trials was considered clinically meaningful. The random‐effects summary was treated as the average of the range of possible treatment effects and we tried to discuss the clinical implications of treatment effects differing between trials. If the average treatment effect was not clinically meaningful, we did not combine trials.

If we used random‐effects analyses, the results were presented as the average treatment effect with 95% CIs, and the estimates of Tau² and I².

Subgroup analysis and investigation of heterogeneity

If we identified substantial heterogeneity, we investigated it using subgroup analyses. We considered whether an overall summary was meaningful, and if it was, used random‐effects analysis to produce it.

Diagnostic tests used

• ADA 2013, IADPSG 2010, Nankervis 2014 versus ACOG 2013 versus NICE 2015 versus NICE 2008; WHO 1999; WHO 2014; Hoffman 1998 versus New Zealand Ministry of Health 2014 versus other not previously specified.

Timing of diagnosis of GDM

• Early (< 28 weeks' gestation) versus late (≥ 28 weeks' gestation)

There was insufficient data for us to be able to explore subgroup analyses for this review. In future updates, if there is sufficient evidence, we plan to use the following outcomes in subgroup analysis.

Maternal outcomes

• Pre‐eclampsia

• Caesarean section

• Development of type 2 diabetes

Neonatal outcomes

• LGA

• Perinatal death or later infant mortality

• Death or serious morbidity composite (variously defined by trials, e.g. infant death, shoulder dystocia, bone fracture or nerve palsy)

• Neurosensory disability in later childhood (as defined by trialists).

We assessed subgroup differences by interaction tests available within RevMan (RevMan 2014). We reported the results of subgroup analyses quoting the Chi² statistic and P value, and the interaction test I² value.

Sensitivity analysis

Where there was evidence of substantial heterogeneity, we explored this by using the quality of the included trials for the primary outcomes. We planned to compare trials that had low risk of bias for allocation concealment with those judged to be of unclear or high risk of bias. We planned to exclude conference abstracts from the meta‐analysis.

In future updates, where data are available we plan to investigate the effect of the randomisation unit (i.e. where included cluster‐randomised trials along with individually‐randomised trials).

Results

Description of studies

Results of the search

We identified 28 potential reports for inclusion in the review (Figure 1). Eleven studies (20 reports) were included and five studies were excluded. Two studies (Coiner 2015; Sheizaf 2006) are awaiting classification and one study (Moore 2016) is ongoing (see Characteristics of studies awaiting classification; Characteristics of ongoing studies).

Included studies

Eleven studies (20 reports) were identified that met the inclusion criteria for this review (Bertini 2005; Casey 2015; Cortez 2006; De Bacco 2015; Fenn 2015; George 2015; Moore 2010; Myers 2014; Nachum 2015; Notelovitz 1971; Silva 2012). Three of these were published in conference abstract format only (Cortez 2006; De Bacco 2015; Nachum 2015).

Design

All 11 studies were randomised parallel design.

Sample sizes

The 11 studies recruited a total of 1487 women and their babies. Sample size ranged from 40 women reported by Myers 2014 to 207 women reported by Notelovitz 1971.

Setting

Three studies were conducted in Brazil (Bertini 2005; De Bacco 2015; Silva 2012) and USA (Casey 2015; Cortez 2006; Moore 2010). Two studies were conducted in India (Fenn 2015; George 2015). One study each was conducted in South Africa (Notelovitz 1971), UK (Myers 2014) and Israel (Nachum 2015).

Participants

Maternal age was reported in six of the 11 studies (Table 8), maternal body mass index (BMI) (kg/m²) was reported in five of 11 studies (Table 9) and gestational age at study entry was reported in four of 11 studies (Table 10).

2. Mean maternal age (years) ± SD.

Study ID	Intervention	Comparison
Bertini 2005	31.2 ± 4.5 (n = 24), glibenclamide	31.5 ± 5.8 (n = 19), acarbose
Casey 2015	31.3 ± 6, glibenclamide	31.2 ± 6, placebo
Cortez 2006	Not stated, acarbose	Not stated, placebo
De Bacco 2015	Not stated, glibenclamide	Not stated, metformin
Fenn 2015	Not stated, glibenclamide	Not stated, metformin
George 2015	33.4 ± 4.4 (n = 79), metformin	33.6 ± 4.6 (n = 80), glibenclamide
Moore 2010	31 ± 7.1 (n = 75) ‐ metformin	29.6 ± 7.8 (n = 74), glibenclamide
Myers 2014	Not stated, metformin	Not stated, standard care
Nachum 2015	Not stated, metformin	Not stated, glibenclamide
Notelovitz 1971	Chlopropramide 30.9 (n = 58) Tolbutamide 29.7 (n = 46)	Diet 32.7 (n = 56)
Silva 2012	32.6 ± 5.6 (n = 104), metformin	31.3 ± 5.4 (n = 96), glibenclamide

3. Maternal BMI kg/m².

Study ID	Intervention	Comparison	Timepoint BMI measured at
Bertini 2005	27.5 ± 5.8 (n = 24), glibenclamide	25.7 ± 4.2 (n = 19), acarbose	Not stated
Casey 2015	29.0 ± 4.8	28.9 ± 5.3	Pre‐pregnancy
Cortez 2006	Not stated	Not stated	Not stated
De Bacco 2015	Not stated	Not stated	Not stated
Fenn 2015	Not stated	Not stated	Not stated
George 2015	28.7 ± 4.4 (n = 79), metformin	28.8 ± 4.0 (n = 80), glibenclamide	Baseline
Moore 2010	32.8 ± 5.8 (n = 75), metformin	32.7 ± 7.0 (n = 74), glibenclamide	Not stated
Nachum 2015	Not stated	Not stated	Not stated
Notelovitz 1971	Not stated	Not stated	Not stated
Silva 2012	28.7 ± 5.4 (n = 104), metformin	28.6 ± 5.9 (n = 96), glibenclamide	Not stated

4. Gestational age at trial entry.

Study ID	Intervention	Comparison
Bertini 2005	Not stated	Not stated
Casey 2015	26.0 ± 2.0, glibenclamide	26.0 ± 1.0, placebo
Cortez 2006	Not stated	Not stated
De Bacco 2015	Not stated	Not stated
Fenn 2015	Not stated	Not stated
George 2015	29.3 ± 3.3 weeks' (n = 79), metformin	29.7 ± 3.7 weeks' (n = 80)
Moore 2010	27.3 ± 6.8 weeks' (n = 75), metformin	29.1 ± 5.0 weeks' (n = 74), glibenclamide
Myers 2014	Not stated	Not stated
Nachum 2015	Not stated	Not stated
Notelovitz 1971	Not stated	Not stated
Silva 2012	27.0 ± 6.4 weeks' (n = 104), metformin	25.4 ± 7.1 weeks' (n = 96), glibenclamide

Interventions and comparisons

Four different comparisons were reported.

Oral anti‐diabetic pharmacological therapy versus placebo or usual care

One study compared glibenclamide with placebo (Casey 2015); one study compared acarbose with placebo (Cortez 2006), one study compared metformin with standard care (Myers 2014) and one study compared tolbutamide and chlorpropamide with diet (Notelovitz 1971).

Metformin versus glibenclamide

Five studies compared metformin with glibenclamide (De Bacco 2015; Fenn 2015; George 2015; Moore 2010; Silva 2012).

Glibenclamide versus acarbose

One study compared glibenclamide with acarbose (Bertini 2005).

Glibenclamide with or without metformin versus metformin with or without glibenclamide

One study compared glibenclamide plus metformin if glycaemic targets were not met, with metformin plus glibenclamide if glycaemic targets were not met (Nachum 2015).

Diagnostic criteria

Four different diagnostic criteria were used by the eight studies that reported this.

• George 2015 and Casey 2015 used the National Diabetes Data Group (NDGG 1979)

• Bertini 2005 and Silva 2012 used WHO (1999) criteria

• Three studies (Fenn 2015; Moore 2010; Nachum 2015) used Carpenter and Coustan criteria

• De Bacco 2015 used WHO criteria but the abstract did not state if this was 1999 or 2015 criteria

• Cortez 2006, Myers 2014 and Notelovitz 1971 did not state which diagnostic criteria they used

Refer to Table 7 and Table 11.

5. Diagnostic criteria.

Study ID	Timing	Screening	Diagnosis	Criteria
Casey 2015	24‐28 weeks'	1 hour, 50 g OGCT(≥ 7.8 mmol/L; 140 mg/dL)	2 abnormal values 3 hour, 100 g OGTT Fasting < 5.8 mmol/L (105 mg/dL) 1‐hour ≥ 10.6 mmol/L (190 mg/dL) 2‐hour ≥ 9.2 mmol/L (165 mg/dL) 3‐hour ≥ 8.1 mmol/L (145 mg/dL)	National Diabetes Data Group
Bertini 2005	11‐33 weeks'	Not stated	75 g OGTT Fasting ≥ 6.1 mmol/L (110 mg/dL); 2‐hour value ≥ 7.8 mmol/L (140 mg/dL).	WHO criteria (old)
Cortez 2006	12‐34 weeks'	Not stated	Not stated	Not stated
De Bacco 2015	Not stated	Not stated	Not stated	WHO criteria but not stated if 1999 or 2015
Fenn 2015	‐	1 hour 50 g OGCT (≥ 7.8 mmol/L; 140 mg/dL)	2 abnormal values 100 g OGTT: fasting glucose ≥ 5.3 mmol/L, 1‐hour ≥ 10 mmol/L, 2‐hour ≥ 8.6 mmol/L, 3‐hour ≥ 7.8 mmol/L	Carpenter and Coustan
George 2015	24‐28 weeks'	Not stated	2 abnormal values 100 g OGTT: fasting glucose ≥ 5.3 mmol/L, 1‐hour ≥ 10 mmol/L, 2‐hour ≥ 8.6 mmol/L, 3‐hour ≥ 7.8 mmol/L	National Diabetes Data Group (1979)
Moore 2010	11‐33 weeks'	1 hour 50 g OGCT (≥ 7.2 mmol/L; 130 mg/dL)	3 hour 100 g OGTT using criteria with 2 or more abnormal results.	Carpenter and Coustan
Myers 2014	Not stated	Not stated	Fasting blood glucose 5.1 to 5.4 mmol/L, 2 hour < 8.5 mmol/L	Not stated
Nachum 2015	11‐33 weeks'	1 hour 50 g OGCT (≥ 7.2 mmol/L; 130 mg/dL)	3 hour 100 g OGTT using criteria with 2 or more abnormal results.	Carpenter and Coustan
Notelovitz 1971	Not stated	Not stated	Not stated	Not stated
Silva 2012	Not stated	Not stated	Not stated	WHO criteria (1999)

OGCT oral glucose tolerance test; OGTT oral glucose tolerance test

Treatment targets

Eight studies reported three different treatment targets for fasting blood glucose (Refer to Table 12):

6. Treatment target.

Study ID	Fasting	1‐hour post‐prandial	2‐hour post‐prandial
Casey 2015	< 5.3 mmol/L (95mg/dL)	‐	< 6.7mmol/L (120 mg/dL)
Bertini 2005	< 5.0 mmol/L (90 mg/dL)	‐	< 5.5 mmol/L (100 mg/dL)
Cortez 2006	< 5.3 mmol/L (95 mg/dL)	< 7.5 mmol/L (135 mg/dL)	‐
De Bacco 2015	Not stated	Not stated	Not stated
Fenn 2015	< 5.3 mmol/L (95 mg/dL)	< 7.8 mmol/L (140 mg/dL)	‐
George 2015	≤ 5.3 mmol/L (95 mg/dL)	‐	≤ 6.7 mmol/L (120 mg/dL)
Moore 2010	< 5.8 mmol/L (105 mg/dL)	‐	< 6.7 mmol/L (120 mg/dL)
Myers 2014	Not stated	Not stated	Not stated
Nachum 2015	≤ 5.3 mmol/L (95 mg/dL)	90 minutes < 7.2 mmol/L (130 mg/dL)	‐
Notelovitz 1971	‐	8.3 mmol/L* (150 mg/dL)	‐
Silva 2012	5.0 mmol/L (90 mg/dL)	< 6.7 mmol/L (120 mg/dL)	‐

Post‐prandial timing not specified

• less than 5.0 mmol/L (90 mg/dL) (Bertini 2005; Silva 2012);

• less than 5.3 mmol/L (95 mg/dL) (Casey 2015; Cortez 2006; Fenn 2015; George 2015; Nachum 2015);

• less than 5.8 mmol/L (105 mg/dL) (Moore 2010).

Four studies reported four different treatment targets for one‐hour postprandial blood glucose:

• less than 6.7 mmol/L (120 mg/dL) (Silva 2012);

• less than 7.2 mmol/L (130 mg/dL) (George 2015) 90 minutes;

• less than 7.5 mmol/L (135 mg/dL) (Cortez 2006);

• less than 7.8 mmol/L (140 mg/dL) (Fenn 2015).

Four studies reported two different treatment targets for two‐hour postprandial blood glucose:

• less than 5.5 mmol/L (100 mg/dL) (Bertini 2005);

• less than 6.7 mmol/L (120 mg/dL) (Casey 2015; George 2015; Moore 2010).

Notelovitz 1971 reported "good control" was obtained with a postprandial (timing not specified) blood glucose reading of less than 8.3 mmol/L (150 mg/dL). De Bacco 2015 and Myers 2014 did not provide details of glycaemic targets for treatment.

The Notelovitz 1971 study included women with both GDM and pre‐gestational diabetes. The proportions of women included with each type of diabetes were unclear and therefore the data have not been included in any meta‐analyses.

See Characteristics of included studies.

Excluded studies

We excluded five studies. One study comparing metformin and placebo was registered with the clinical studies registry ClinicalTrials.gov in 2010. Subsequent updates indicate that the study never started recruitment due to insufficient funding for enrolment of participants (Branch 2010). Hebert 2011, was a pharmacokinetic study and not an interventional study. We excluded Ainuddin 2013 as the comparison was ineligible for inclusion in this review, the trial compared metformin and insulin. We excluded two studies as the intervention was postpartum (Berens 2015; Smith 2015). See Characteristics of excluded studies.

Risk of bias in included studies

Refer to Figure 2; Figure 3 and Characteristics of included studies for 'risk of bias' summaries of the included studies.

2.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies

3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study

Allocation

We judged five of the 11 included studies to be of low risk of bias for random sequence generation as they all used computer‐generated number tables (Casey 2015; Fenn 2015; George 2015; Moore 2010; Silva 2012). The remaining studies stated that they were randomised but did not provide sufficient details to make a judgement so we therefore considered them to be unclear risk of bias (Bertini 2005; Cortez 2006; De Bacco 2015; Myers 2014; Nachum 2015; Notelovitz 1971).

We considered allocation concealment to be low risk of bias in five studies. Bertini 2005 reported using "brown envelopes containing outside the randomisation number and in the inside a sheet defining which therapy the patient was allocated to"; Casey 2015 reported that masking, allocation and assignment was done by the investigational drug pharmacy; George 2015 used "sequentially labelled opaque envelopes" that were maintained "in a central research office by research officers not involved in patient care". Moore 2010 used "sequentially labelled, opaque, sealed envelopes" and Silva 2012 used sequential numbering in brown envelopes. Myers 2014 reported using "..prefilled sealed envelopes created by independent research midwives within the department". The remaining studies provided no details as to the method of allocation concealment and we judged them to be of unclear risk of bias (Cortez 2006; De Bacco 2015; Fenn 2015; Nachum 2015; Notelovitz 1971).

Blinding

Performance bias

Casey 2015 reported that the "subject", "caregiver" and "investigator" were blinded and the study was placebo controlled (we obtained additional information from the study protocol). Sham dose adjustments were made in the placebo group. Fenn 2015 reported that the physician and the participant were blinded to treatment allocation and that medication was provided from the pharmacy in a sealed envelope. We judged these studies to be low risk of bias.

Cortez 2006 reported that the study was double‐blind but gave no details as to who was blinded. We judged the study to be of unclear risk of bias.

Six studies reported that they were open label and therefore participants and staff were not blinded to the treatment allocation (Bertini 2005; De Bacco 2015; Moore 2010; Myers 2014; Nachum 2015; Silva 2012). George 2015 reported that the women in their study were not blinded to the treatment allocation. We considered these studies to be of high risk of bias.

Notelovitz 1971 did not provide any details regarding blinding and we therefore judged it as being unclear risk of bias.

Detection bias

In the trial registration document, Casey 2015 reported that the "outcome assessor" was blinded. George 2015 reported that the neonatologists who cared for the infants after birth were blinded to treatment allocation. We judged these studies to be low risk of bias.

Cortez 2006 reported that the study was double‐blind but gave no details as to who was blinded. We judged the study to be of unclear risk of bias.

Eight studies reported that they were open label or provided no details relating to blinding of outcome assessors, or both (Bertini 2005; De Bacco 2015; Fenn 2015; Moore 2010; Myers 2014; Nachum 2015; Notelovitz 1971; Silva 2012). We considered these studies to be of unclear risk of bias.

Incomplete outcome data

There was no evidence of risk of attrition bias in six studies (Bertini 2005; Fenn 2015; George 2015; Moore 2010; Notelovitz 1971; Silva 2012). Casey 2015 reported a 5% (20/395) loss to follow‐up but did not provide any reasons for the attrition. No details of attrition were provided by Cortez 2006, although the study stated that analysis was conducted by intention to treat. We judged this study to be of unclear risk of bias. In De Bacco 2015 8/36 (22%) women dropped out of the metformin arm and 24/45 (55%) dropped out of the glibenclamide arm. We judged this study to be high risk of bias. In the Myers 2014 pilot study, three women did not complete the trial and were not analysed, but it is unclear as to which group they were allocated ‐ we assessed this study to be at an unclear risk of bias.

Selective reporting

There was no evidence of risk of reporting bias in three studies (Bertini 2005; George 2015; Moore 2010).

Casey 2015 reported additional maternal and neonatal outcomes to those listed a priori in the trial registration document. Three studies were only published in abstract form and or no full publication could be identified (Cortez 2006; De Bacco 2015; Myers 2014). One study failed to report on an outcome that had been prespecified (Silva 2012), two studies reported additional outcomes that were not pre‐specified (Fenn 2015; Nachum 2015) and one study did not pre‐specify any outcomes (Notelovitz 1971). We judged these studies to be of high risk of bias.

Other potential sources of bias

There was no evidence of other sources of bias in three studies (Bertini 2005; Moore 2010; Silva 2012).

Fenn 2015 did not provide data for the population demographics and we were unable to judge the risk of bias, which we have allocated an unclear risk.

In one study (George 2015) an interim analysis requested by the local data monitoring committee showed significant differences in outcomes and the study was stopped before the total sample size of 86 women per group was achieved. The intervention and control groups were balanced at baseline although the metformin group had higher fasting triglyceride levels. We judged the study to be of unclear risk of bias.

Three studies were reported in abstract form only and there were no baseline demographic data (Cortez 2006; De Bacco 2015; Nachum 2015). The Myers 2014 study remains unpublished. We therefore judged these studies to be of high risk of bias.

The Casey 2015 study appears to have been registered twice as NCT00942552 and as NCT00744965 with the same outcomes, interventions and sample size. The population was 93% Hispanic and therefore the results may not be generalisable to other ethnicities. We judged the study to be high risk of bias.

We could not separate the data from the Notelovitz 1971 study for women with GDM and those with pregestational diabetes, and the study did not report the proportions of women for these categories. We did not, therefore, include the data in any meta‐analyses.

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4; Table 5; Table 6

Oral anti‐diabetic pharmacological therapy versus placebo or standard care (comparison 1)

Three studies reported data for this comparison (Casey 2015; Cortez 2006; Myers 2014). Casey 2015 compared glibenclamide with placebo and Cortez 2006 compared acarbose with placebo. Myers 2014 compared metformin with standard care.

Maternal primary outcomes

1.1 Hypertensive disorders of pregnancy

One study reported on hypertensive disorders of pregnancy in a study that compared glibenclamide with a placebo (Casey 2015). Overall, for hypertensive disorders of pregnancy (any type) there was no difference between glibenclamide and placebo groups (risk ratio (RR) 1.24, 95% confidence interval (CI) 0.81 to 1.90; one study, n = 375 women; Analysis 1.1). Using GRADE the quality of the evidence was judged to be very low due to no published protocol being identified; more outcomes being reported than were listed in the trial registration documentation; evidence of indirectness, as 93% of population were Hispanic and evidence may not be generalisable; and publication bias, as the evidence was only based on a single study. The risk of developing a hypertensive disorder of pregnancy in the placebo group was 16.7%, if the woman had been treated with glibenclamide the risk would range from 13.5% to 31.7% (Table 1).

1.1. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 1 Hypertensive disorders of pregnancy.

For pregnancy‐induced hypertension (persistent systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg) there was no difference between glibenclamide and placebo groups (RR 1.24, 95% CI 0.71 to 2.19; one study, n = 375 women; Analysis 1.1 ). For severe pregnancy‐induced hypertension (proteinuria ≥ 2 g in 24 hours, or ≥ 2+ on dipstick, blood pressure ≥ 160 mmHg or diastolic blood pressure ≥ 110 mmHg, serum creatinine > 1.0 mg/dL, platelets < 100,000/mm³, aspartate aminotransferase > 90 units/L, or symptoms such as persistent headache, scotomata or epigastric pain) there was no difference between glibenclamide and placebo groups (RR 1.23, 95% CI 0.59 to 2.56; one study, 375 women; Analysis 1.1).

1.2 Caesarean section

One study reported on caesarean section in a study that compared glibenclamide with a placebo (Casey 2015). There was no difference in the risk for birth by caesarean section between glibenclamide and placebo groups (RR 1.03, 95% CI 0.79 to 1.34; one study, n = 375 women; Analysis 1.2). Using GRADE the quality of the evidence was judged to be very low due to no published protocol being identified; more outcomes being reported than were listed in the trial registration documentation; evidence of indirectness, as 93% of population were Hispanic and evidence may not be generalisable; and publication bias, as the evidence was only based on a single study. The risk of birth by caesarean section in the placebo group was 36%. For women treated with glibenclamide the risk would range from 28.5% to 48.3% (Table 1).

1.2. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 2 Caesarean section.

Development of type 2 diabetes

None of the included trials comparing oral anti‐diabetic pharmacological therapies with placebo pre‐specified development of type 2 diabetes as a study outcome.

Neonatal primary outcomes

1.3 LGA (90th percentile and above)

There was no difference in the risk of being born LGA age between infants whose mothers had been treated with glibenclamide and those who had received placebo (RR 0.89, 95% CI 0.51 to 1.58; one study, 375 infants; Analysis 1.3) (Casey 2015). Using GRADE the quality of the evidence was judged to be low due to no published protocol being identified; more outcomes being reported than were listed in the trial registration documentation; evidence of indirectness, as 93% of population were Hispanic and evidence may not be generalisable; and publication bias as the evidence was only based on a single study. The risk of being born LGA if the mother had been treated with placebo was 11.8%, if the mother had been treated with glibenclamide the risk would have ranged from 6% to 18.7%.

1.3. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 3 Large‐for‐gestational age.

Other neonatal primary outcomes

None of the included trials comparing oral anti‐diabetic pharmacological therapies with placebo pre‐specified perinatal death and later infant mortality, death or serious morbidity composite or neurosensory disability in later childhood as study outcomes.

Maternal secondary outcomes

1.4 Use of additional pharmacotherapy

There was no evidence of a difference in the need for insulin between the anti‐diabetic pharmacological therapy and placebo groups reported in two studies (Casey 2015; Cortez 2006) (RR 0.68, 95% CI 0.42 to 1.11; two studies, n = 434 women; Analysis 1.4). The Casey 2015 study used glibenclamide and the Cortez 2006 study used acarbose as the pharmacological intervention. Myers 2014 reported that 15/19 women in the standard care group were prescribed metformin and two women required insulin. It is not clear if any of the women in the metformin group required supplementary insulin.

1.4. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 4 Use of additional pharmacotherapy.

1.5 Glycaemic control during/end of treatment

Glibenclamide was associated with a reduction in fasting capillary glucose concentrations taken at the last three antenatal visits compared with placebo (mean difference (MD) ‐3.00 mg/dL, 95% CI ‐5.13 to ‐0.87; one study, n = 375 women; Analysis 1.5) (Casey 2015). Caution is required in interpreting these data from a single trial with evidence of imprecision.

1.5. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 5 Glycaemic control (end of treatment) (mg/dL).

1.6 Weight gain in pregnancy

There was no evidence of a difference between glibenclamide and placebo groups for weight gain during pregnancy (MD 0.00 Kg, 95% CI ‐0.96 to 0.96; one study, n = 375 women; Analysis 1.6) (Casey 2015). Caution is required in interpreting these data from a single trial with evidence of imprecision.

1.6. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 6 Weight gain in pregnancy (Kg).

1.7 Induction of labour

There was no evidence of a difference between glibenclamide and placebo groups for the risk of induction of labour reported in one study (Casey 2015) (RR 1.18, 95% CI 0.79 to 1.76; one study, n = 375 women; Analysis 1.7). Using GRADE the quality of the evidence was judged to be very low due to no published protocol being identified; more outcomes being reported than were listed in the trial registration documentation; evidence of indirectness, as 93% of population were Hispanic and therefore the evidence may not be generalisable; and publication bias, as the evidence was only based on a single study. The risk for induction of labour in the placebo group was 18.8%, if the woman was treated with glibenclamide the risk would range from 14.9% to 33.1% (Table 1).

1.7. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 7 Induction of labour.

1.8 Perineal trauma/tearing

There was no evidence of a difference between glibenclamide and placebo groups for the risk of perineal trauma reported as third‐ to fourth‐degree tear reported in one study (Casey 2015) (RR 0.98, 95% CI 0.06 to 15.62; one study, n = 375 women; Analysis 1.8). Using GRADE the quality of the evidence was judged to be very low due to no published protocol being identified; more outcomes being reported than were listed in the trial registration documentation; evidence of indirectness, as 93% of population were Hispanic and evidence may not be generalisable; and wide CIs crossing the line of no effect and low event rates were suggestive of imprecision and publication bias, as the evidence was only based on a single study. If the risk of perineal trauma in the placebo group was 0.5%, then between 0% to 8% of the women treated with glibenclamide would be at risk of perineal trauma (Table 1).

1.8. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 8 Perineal trauma.

Other maternal secondary outcomes

No data were reported for any of the other pre‐specified maternal secondary outcomes for this review (maternal hypoglycaemia, adherence to the intervention, placental abruption, postpartum haemorrhage, breastfeeding at discharge, six weeks postpartum, six months or longer, maternal mortality, sense of well‐being and quality of life, behavioural changes associated with the intervention, views of the intervention, relevant biomarker changes associated with the intervention).

Long‐term maternal outcomes

No data were reported for long‐term maternal outcomes (post‐natal depression, BMI, postnatal weight retention or return to pre‐pregnancy weight, type 1 diabetes, type 2 diabetes, impaired glucose tolerance, subsequent GDM, cardiovascular health (as defined by trialists including blood pressure, hypertension, cardiovascular disease, metabolic syndrome).

Neonatal secondary outcomes

1.9 Stillbirth

One study reported on stillbirth (Casey 2015). There was no evidence of a difference in the risk of stillbirth between the glibenclamide and placebo groups (RR 0.49, 95% CI 0.05 to 5.38; one study, n = 375 infants; Analysis 1.9). Using GRADE the quality of the evidence was judged to be very low. This was due to no published protocol being identified; more outcomes being reported than were listed in the trial registration documentation; evidence of indirectness, as 93% of population were Hispanic and evidence may not be generalisable; and wide CIs crossing the line of no effect and low event rates were suggestive of imprecision and publication bias, as the evidence was only based on a single study. If the risk of stillbirth was 1.1% for the placebo group, between 0.1% to 5.8% of infants would have been stillborn if the woman was treated with glibenclamide.

1.9. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 9 Stillbirth.

1.10 Neonatal death

There were no events of neonatal death reported in either the glibenclamide or the placebo group in one study including 375 infants (Casey 2015).

1.11 Small‐for‐gestational age

There was no evidence of a difference in the risk of being born small‐for‐gestational age between the glibenclamide and placebo groups reported in one study (Casey 2015) (RR 1.11, 95% CI 0.58 to 2.10; one study, n = 375 infants; Analysis 1.11).

1.11. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 11 Small‐for‐gestational age.

1.12 Macrosomia

There was no evidence of a difference in the risk of macrosomia (≥ 4 kg) between infants whose mothers had been treated with glibenclamide and those who had received placebo reported in one study (Casey 2015) (RR 0.71, 95% CI 0.36 to 1.41; one study, n = 375 infants; Analysis 1.12).

1.12. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 12 Macrosomia.

1.13 Birthweight

There was no evidence of a difference in birthweight between infants whose mothers had been treated with glibenclamide and those who had received placebo reported in one study (Casey 2015) (MD ‐33.00 g, 95% CI ‐134.53 to 68.53; one study, n = 375 infants).

1.14 to 1.16 Birth trauma (shoulder dystocia, bone fracture, nerve palsy)

Birth trauma was reported in one study (Casey 2015) comparing glibenclamide with placebo. There was no evidence of a difference in the risk for shoulder dystocia (RR 0.33, 95% CI 0.01 to 8.00; one study, n = 375 infants; Analysis 1.14); bone fracture (RR 0.74, 95% CI 0.17 to 3.25; one study, n = 375 infants; Analysis 1.15) or nerve palsy (RR 0.33, 95% CI 0.01 to 8.00; one study, n = 375; Analysis 1.15) between the glibenclamide and placebo groups. Caution is required when interpreting these data due to low event rates and wide CIs crossing the line of no effect suggesting imprecision.

1.14. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 14 Shoulder dystocia.

1.15. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 15 Bone fracture.

1.17 Gestational age at birth

There was no evidence of a difference between glibenclamide and placebo groups for the gestational age at birth (MD 0.00 weeks, 95% CI ‐0.32 to 0.32; one study, n = 375 infants; Analysis 1.17) (Casey 2015).

1.17. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 17 Gestational age at birth (weeks).

1.18 Neonatal hypoglycaemia

There was no evidence of a difference in the risk for neonatal hypoglycaemia between the glibenclamide and the placebo groups (RR 1.97, 95% CI 0.36 to 10.62; one study, n = 375 infants; Analysis 1.18) (Casey 2015). Caution is required when interpreting these data due to low event rates and wide CIs crossing the line of no effect suggesting imprecision. Using GRADE the quality of the evidence was judged to be very low due to no published protocol being identified; more outcomes being reported than were listed in the trial registration documentation; evidence of indirectness, as 93% of population were Hispanic and evidence may not be generalisable; and wide CIs crossing the line of no effect and low event rates were suggestive of imprecision and publication bias, as the evidence was only based on a single study. If the risk of neonatal hypoglycaemia in the placebo group was 11%, between 4% to 11.4% of infants would experience hypoglycaemia if the woman was treated with glibenclamide.

1.18. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 18 Neonatal hypoglycaemia.

1.19 Neonatal jaundice (hyperbilirubinaemia)

There was no evidence of a difference in the risk for neonatal jaundice between the glibenclamide and the placebo groups reported in one study (RR 1.97, 95% CI 0.50 to 7.75; one study, n = 375 infants; Analysis 1.19) (Casey 2015).

1.19. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 19 Hyperbilirubinaemia.

Other neonatal secondary outcomes

No data were reported for any of the other neonatal secondary outcomes for this review (preterm birth (less than 37 weeks’ gestation; and less than 32 weeks' gestation), five‐minute Apgar less than seven, birthweight z score, head circumference and z score, length and z score, ponderal index, adiposity, respiratory distress syndrome, hypocalcaemia, polycythaemia, relevant biomarker changes associated with the intervention).

Later infant/childhood outcomes

No data were reported for childhood outcomes (weight and z scores, height and z scores, head circumference and z scores, adiposity, educational attainment, blood pressure, type 1 diabetes, type 2 diabetes, impaired glucose tolerance, dyslipidaemia or metabolic syndrome).

Child as an adult outcomes

No data were reported for child as an adult outcomes (weight, height, adiposity, cardiovascular health (as defined by trialists including blood pressure, hypertension, cardiovascular disease, metabolic syndrome), employment, education and social status/achievement, dyslipidaemia or metabolic syndrome, type 1 diabetes, type 2 diabetes, impaired glucose tolerance).

Health service outcomes

1.20 Admission to NICU

There was no evidence of a difference in the risk of admission to neonatal intensive care between the glibenclamide and the placebo groups reported in one study (Casey 2015) (RR 1.16, 95% CI 0.53 to 2.53; one study, n = 375 infants; Analysis 1.20).

1.20. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 20 Admission to NICU.

Other health service outcomes

No data were reported for the remaining health service outcomes (number of antenatal visits or admissions, number of hospital or health professional visits (including midwife, obstetrician, physician, dietician, diabetic nurse), length of antenatal stay, length of postnatal stay (maternal), length of postnatal stay (baby), cost of maternal care, cost of offspring care, costs associated with the intervention, costs to families associated with the management provided, cost of dietary monitoring (e.g. diet journals, dietician, nurse visits, etc), costs to families ‐ change of diet, extra antenatal visits, extra use of healthcare services (consultations, blood glucose monitoring, length and number of antenatal visits), women’s view of treatment advice, duration of stay in neonatal intensive care unit or special care baby unit).

Metformin versus glibenclamide (comparison 2)

Six studies compared metformin with glibenclamide (De Bacco 2015; Fenn 2015; George 2015; Moore 2010; Nachum 2015; Silva 2012).

Maternal primary outcomes

2.1 Hypertensive disorders of pregnancy

Overall there was no evidence of a difference between metformin and glibenclamide‐treated groups for the risk of hypertensive disorders of pregnancy (any definition) (RR 0.70, 95% CI 0.38 to 1.30; three studies, n = 508 women; Analysis 2.1). Using GRADE the quality of the evidence was judged to be moderate, due to the trials being open‐label. If the risk of developing a hypertensive disorder of pregnancy was 8.8% in the glibenclamide group; for those treated with metformin the risk would range from 3.3% to 11.4%.

2.1. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 1 Hypertensive disorders of pregnancy.

One study reported data for pre‐eclampsia (not defined) (Moore 2010). There was no evidence of a difference in the risk for pre‐eclampsia (not defined) between women who had been treated with metformin and those who had been treated with glibenclamide (RR 0.66, 95% CI 0.11 to 3.82; one study, n = 149 women). Two studies reported on pregnancy‐induced hypertension, George 2015 did not provide a definition and Silva 2012 reported on the presence of chronic arterial systemic hypertension (no further details). There was no difference in the risk for pregnancy‐induced hypertension between women who had been treated with metformin and those who had been treated with glibenclamide (RR 0.71, 95% CI 0.37 to 1.37; two studies, n = 359 women (Analysis 2.1).

2.2 Caesarean section

Four studies reported data for caesarean section (Fenn 2015; George 2015; Moore 2010; Silva 2012). There was no evidence of a difference in the risk for birth by caesarean section between women who had been treated with metformin and those who had been treated with glibenclamide (average RR 1.20, 95% CI 0.83 to 1.72; four studies, n = 554 women; I² = 61%, Tau² = 0.07) and the heterogeneity could not be explained through the diagnostic criteria used (Analysis 2.2). We excluded the study by Silva 2012 as this had unclear risk of bias for allocation concealment; heterogeneity increased to I² = 69%, Tau² = 0.11. The results of the Moore 2010 study, which included primarily Hispanic women with a mean BMI greater than 30 kg/m², appeared to differ from the other studies and, when removed from the meta‐analysis, heterogeneity was reduced from I² = 61% to I² = 0% (RR 1.01, 95% CI 0.86 to 1.20; three studies, n = 405 women). The lack of difference between the metformin and glibenclamide groups for the chance of birth by caesarean section remained unaltered by the removal of Moore 2010. Using GRADE the overall quality of the evidence was low due to lack of blinding and substantial heterogeneity. If the risk of birth by caesarean section when the woman had been treated with glibenclamide was 39.2%, the risk if the woman had been treated with metformin would have ranged from 32.5% to 67.4%.

2.2. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 2 Caesarean section.

Development of type 2 diabetes

No data were reported for development of type 2 diabetes in any of the included studies for this comparison.

Neonatal primary outcomes

2.3 Perinatal death

Two studies reported data for perinatal death, George 2015 reported no events in either the metformin or the glibenclamide groups; Silva 2012 reported one stillbirth in each group (RR 0.92, 95% CI 0.06 to 14.55; two studies, n = 359 infants) (Analysis 2.3). Using GRADE the quality of the evidence was considered to be very low due to lack of blinding, imprecision and low event rates. If the risk of perinatal death when the mother had been treated with glibenclamide was 6%, the risk if the mother had been treated with metformin would have ranged from 0% to 8.3%.

2.3. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 3 Perinatal mortality.

2.4 LGA

There was no evidence of a difference in the risk of being born LGA when the mother was treated with metformin compared with glibenclamide reported in two studies (Fenn 2015; Silva 2012) (average RR 0.67, 95% CI 0.24 to 1.83; two studies, n = 246 infants; random‐effects model, I² = 54%, Tau² = 0.30) (Analysis 2.4). Both studies defined LGA as being greater than 90th percentile in growth curves. The subgroup interaction test for diagnostic criteria used was not significant suggesting that there was no differential effect based on diagnostic criteria. Using GRADE the quality of the evidence was considered to be low due to imprecision and heterogeneity. If the risk of being born LGA if the mother had been treated with glibenclamide was 19%, the risk if the mother had been treated with metformin would have ranged from 4.6% to 35.4%.

2.4. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 4 Large‐for‐gestational age.

2.5 Death or serious morbidity composite

One study (George 2015) reported data for a composite of neonatal outcomes that included hypoglycaemia, hyperbilirubinaemia, macrosomia, respiratory illness, birth injury, stillbirth or neonatal death. Metformin was associated with a reduction in the risk of the composite outcome compared with glibenclamide (RR 0.54, 95% CI 0.31 to 0.94; one study, n = 159 infants; Analysis 2.5). The evidence should be interpreted with caution as it is based on a single small study. Using GRADE the quality of the evidence was judged to be low due to lack of blinding and publication bias. If the risk of mortality or serious neonatal mortality was 35% if the mother had been treated with glibenclamide, the risk if the mother had been treated with metformin would have ranged from 10.9% to 32.9%.

2.5. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 5 Death or serious morbidity composite.

Neurosensory disability in later childhood

No data were reported for neurosensory disability in later childhood for this comparison.

Maternal secondary outcomes

2.6 Use of additional pharmacotherapy

Five studies reported on the need for additional pharmacological therapy (Fenn 2015; George 2015; Moore 2010; Nachum 2015; Silva 2012). There was no evidence of a difference in the risk of requiring supplemental insulin between women who had been treated with metformin and those treated with glibenclamide (RR 0.66, 95% CI 0.28 to 1.57; five studies, n = 660 women; random‐effects model, I² = 72%, Tau² = 0.57; Analysis 2.6). The heterogeneity could not be explained through the diagnostic criteria used (data not shown). In sensitivity analysis we removed two studies which had unclear risk of bias for allocation concealment (Fenn 2015; Nachum 2015); heterogeneity was not affected by their removal (I² = 77%, Tau² = 0.50) (data not shown). The Moore 2010 study again appeared to have data that differed from the other studies included in the meta‐analysis and exclusion of this study reduced heterogeneity to I² = 8% but also suggested that the use of metformin was associated with a reduced requirement for additional pharmacotherapy (RR 0.56, 95% CI 0.33 to 0.95; four studies, n = 511 women). The authors of the Moore 2010 study suggested that their findings may differ from those of other studies as their study population comprised Hispanic women and they believe there may be an ethnically differential response to metformin in these women.

2.6. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 6 Use of additional pharmacotherapy.

2.7 Maternal hypoglycaemia

Three studies reported on maternal hypoglycaemia. Fenn 2015 and George 2015 did not provide any definitions of maternal hypoglycaemia and Moore 2010 used a cut off of 3.3 mmol/L (60 mg/dL). There was no difference in the risk of maternal hypoglycaemia between women who had been treated with metformin and those treated with glibenclamide (RR 0.89, 95% CI 0.36 to 2.19; three studies, n = 354 women; Analysis 2.7). The De Bacco 2015 study, currently published as a conference abstract, reported data for women who dropped out of the study with hypoglycaemia. Thirty‐nine percent of women in the glibenclamide group (17/45) and 3% of women in the metformin group (1/36) "dropped out" of the study due to maternal hypoglycaemia. It is unclear whether they withdrew from the study due to treatment side effects or were lost to follow‐up.

2.7. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 7 Maternal hypoglycaemia.

2.8 Glycaemic control during/at end of treatment

Three studies reported fasting and postprandial blood glucose concentrations (Moore 2010; Silva 2012; George 2015) and one study (Silva 2012) reported HbA1c concentration .

Metformin was associated with an increase in fasting blood glucose concentration compared with glibenclamide (standardised mean difference (SMD) 0.19, 95% CI 0.02 to 0.37; three studies, n = 508 women; Analysis 2.8). There was no difference between metformin and glibenclamide for two‐hour (after dinner, where reported) post‐prandial blood glucose concentration (SMD 0.16, 95% CI ‐0.01 to 0.34; three studies, n = 508 women; Analysis 2.8). The data for Silva 2012 are for one‐hour postprandial (mealtime not specified).

2.8. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 8 Glycaemic control (mg/L; mmol/L).

One study (Silva 2012) found no evidence of a difference in HbA1c concentration in the third trimester of the pregnancy between treatment with metformin and treatment with glibenclamide (SMD ‐0.12 %, 95% CI ‐0.39 to 0.16; one study, n = 200 women). The De Bacco 2015 study, currently published as a conference abstract, reported data for women who "dropped out" of the study as they were unable to attain glycaemic control. Fourteen per cent of the women in the glibenclamide group (6/45) and 3% in the metformin group (1/36) dropped out because they were unable to maintain glycaemic control.

2.9 Weight gain in pregnancy

Metformin was associated with a reduced weight during pregnancy compared with glibenclamide, reported by one study (Silva 2012) (MD ‐2.06 Kg, 95% CI ‐3.98 to ‐0.14; one study, n = 200 women; Analysis 2.9).

2.9. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 9 Weight gain in pregnancy (Kg).

2.10 Induction of labour

One study (George 2015) found no evidence of a difference in the risk for induction of labour between women who had been treated with metformin and those treated with glibenclamide (RR 0.81, 95% CI 0.61 to 1.07; one study; n = 159 women; Analysis 2.10). Using GRADE the quality of the evidence was judged to be low due to lack of blinding and publication bias. If the risk for induction of labour if treated with glibenclamide was 61.3%, the risk if treated with metformin would have ranged from 37.4% to 65.5%.

2.10. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 10 Induction of labour.

2.11 Perineal trauma/tearing

There was no evidence of a difference in the risk for third or fourth degree perineal tearing between women who had been treated with metformin and those treated with glibenclamide reported in two studies (George 2015; Moore 2010) (RR 1.67, 95% CI 0.22 to 12.52; two studies, n = 308 women; Analysis 2.11). Using GRADE the quality of the evidence was judged to be low due to lack of blinding, imprecision and low event rates. If the risk of perineal trauma for women treated with glibenclamide was 0.6%, for those treated with metformin the risk would have ranged from 0.1% to 8%.

2.11. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 11 Perineal trauma.

Other maternal secondary outcomes

No data were reported for any of the remaining maternal secondary outcomes for this review (Adherence to the intervention, placental abruption, postpartum haemorrhage, postpartum infection, breastfeeding at discharge, six weeks postpartum, six months or longer, maternal mortality, sense of well‐being and quality of life, behavioural changes associated with the intervention, views of the intervention, relevant biomarker changes associated with the intervention).

Long‐term maternal outcomes

No data were reported for long‐term maternal outcomes (postnatal depression, BMI, postnatal weight retention or return to pre‐pregnancy weight, type 1 diabetes, type 2 diabetes, impaired glucose tolerance, subsequent GDM, cardiovascular health (as defined by trialists including blood pressure, hypertension, cardiovascular disease, metabolic syndrome).

Neonatal secondary outcomes

2.12 Stillbirth

One study (Silva 2012) found no evidence of a difference in the risk for fetal death for infants whose mothers had been treated with metformin or those treated with glibenclamide (RR 0.92, 95% CI 0.06 to 14.55; one study, n = 200 infants; Analysis 2.12).

2.12. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 12 Stillbirth.

2.13 Macrosomia

Macrosomia was defined as birthweight of 3.7 kg or more by George 2015 and as 4.0 kg or more by Moore 2010. There was no evidence of a difference in the risk for macrosomia for infants whose mothers had been treated with metformin or those treated with glibenclamide (RR 0.72, 95% CI 0.23 to 2.21; two studies, n = 308 infants; Analysis 2.13).

2.13. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 13 Macrosomia.

2.14 to 2.15 Birth trauma; shoulder dystocia

One study reported no events of birth injury (no further details) in either the metformin‐ or the glibenclamide‐treated groups (George 2015). Two studies (Fenn 2015; Moore 2010) reported no evidence of a difference in the risk for shoulder dystocia for infants whose mothers had been treated with metformin or those treated with glibenclamide (RR 0.99, 95% CI 0.14 to 6.89; two studies, n = 195 infants) (Analysis 2.15).

2.15. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 15 Shoulder dystocia.

2.16 Gestational age at birth

There was no evidence of a difference in gestational age at birth for infants whose mothers had been treated with metformin or those treated with glibenclamide reported in three studies (George 2015; Moore 2010; Silva 2012) (MD 0.03 weeks, 95% CI ‐0.22 to 0.28; three studies, n = 508 infants; Analysis 2.16). This is probably a reflection of local policies for timing of birth for women with GDM.

2.16. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 16 Gestational age at birth (weeks).

2.17 Preterm birth (less than 37 weeks’ gestation);

There was no evidence of a difference in the risk for preterm birth (less than 37 weeks' gestation) for infants whose mothers had been treated with metformin or those treated with glibenclamide reported in three studies (George 2015; Moore 2010; Silva 2012) (RR 1.59, 95% CI 0.59 to 4.29; three studies, n = 508 infants; Analysis 2.17).

2.17. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 17 Preterm birth.

2.18 Five‐minute Apgar less than seven

There were no events of a five‐minute Apgar score less than seven in either the metformin‐ or glibenclamide‐treated groups in the single study reporting this outcome (Moore 2010). Silva 2012 reported no difference in Apgar scores less than seven with 5/104 events for the metformin group and 5/96 events for the glibenclamide group. However, they do not specify if this is at the one minute or five‐minute assessment.

2.19 Birthweight

Three studies reported on birthweight. George 2015 reported data adjusted for gestational age, maternal height and gender (metformin 3064 g ± 202 versus glibenclamide 3037 g ± 204). We did not include data from this study in the meta‐analysis because heterogeneity was I² = 86% when we included the study and I² = 0% when we removed it. We have contacted the study authors in an attempt to obtain the unadjusted data for birthweight. If these data are available we will add them to the meta‐analysis in the next update of this review.

For the remaining two studies (Moore 2010; Silva 2012), metformin was associated with a reduction in birthweight compared to glibenclamide (MD ‐209.13 g, 95% CI ‐314.53 to ‐103.73; two studies, I² = 0% n = 249 infants; Analysis 2.19).

2.19. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 19 Birthweight (g).

2.20 Ponderal index

Metformin was associated with a reduced ponderal index compared with glibenclamide reported by Silva 2012 (MD ‐0.09 units, 95% CI ‐0.17 to ‐0.01; one study, n = 200 infants; Analysis 2.20).

2.20. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 20 Ponderal index.

2.21 Neonatal hypoglycaemia

There was no evidence of a difference in the risk for neonatal hypoglycaemia (defined as less than 2.2 mmol/L; 40 mg/dL) in infants whose mothers had been treated with metformin or those treated with glibenclamide (RR 0.86, 95% CI 0.42 to 1.77; four studies, n = 554 infants; Analysis 2.21). Using GRADE the quality of the evidence was judged to be low due to lack of blinding and low event rates suggesting imprecision. If the risk of neonatal hypoglycaemia in the group treated with glibenclamide was 4.8%, for those treated with metformin the risk ranged from 1.9% to 8.3%.

2.21. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 21 Neonatal hypoglycaemia.

2.22 Respiratory distress syndrome

There was no evidence of a difference in the risk for neonatal hypoglycaemia in infants whose mothers had been treated with metformin or those treated with glibenclamide reported by George 2015 (RR 0.51, 95% CI 0.10 to 2.69; one study, n = 159 infants; Analysis 2.22).

2.22. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 22 Respiratory distress syndrome.

2.23 Neonatal jaundice (hyperbilirubinaemia)

There was no evidence of a difference in the risk for hyperbilirubinaemia in infants whose mothers had been treated with metformin or those treated with glibenclamide reported by Fenn 2015 and George 2015 (RR 0.68, 95% CI 0.37 to 1.25; two studies, n = 205 infants; Analysis 2.23).

2.23. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 23 Hyperbilirubinaemia.

Relevant biomarkers associated with the intervention

Cord C‐peptide greater than 1.7 micrograms per litre was reported in eight babies whose mothers had been treated with metformin and in five babies whose mothers had been treated with glibenclamide in one study (Fenn 2015).

Other outcomes not prespecified

The De Bacco 2015 study, currently published as a conference abstract, reported data for women who dropped out of the study due to gastric intolerance. Two per cent of women (1/45) in the glibenclamide group and 17% in the metformin group (6/36) withdrew from the study due to gastric intolerance.

Other secondary neonatal outcomes

No data were reported for the other neonatal secondary outcomes for this comparison (neonatal death, small‐for‐gestational age; bone fracture; nerve palsy; congenital malformation; head circumference and z scores; length and z scores; birthweight z scores, head circumference z scores, skinfold thickness measurements (mm); fat mass; hypocalcaemia; hypercalcaemia; polycythaemia).

Later infant/childhood outcomes

No data were reported for childhood outcomes (weight and z scores, height and z scores, head circumference and z scores, adiposity, educational attainment, blood pressure, type 1 diabetes, type 2 diabetes, impaired glucose tolerance, dyslipidaemia or metabolic syndrome).

Child as an adult outcomes

No data were reported for child as an adulthood outcomes (weight, height, adiposity, cardiovascular health (as defined by trialists including blood pressure, hypertension, cardiovascular disease, metabolic syndrome), employment, education and social status/achievement, dyslipidaemia or metabolic syndrome, type 1 diabetes, type 2 diabetes, impaired glucose tolerance).

Health service outcomes

2.24 Admission to neonatal intensive care

There was no evidence of a difference for admission to the neonatal intensive care unit between the metformin‐ and glibenclamide‐treated groups reported by Moore 2010 and Silva 2012 (RR 1.52, 95% CI 0.65 to 3.56; two studies, n = 349 infants; Analysis 2.24).

2.24. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 24 Admission to NICU.

Other health service outcomes

No data were reported for other health service outcomes (number of antenatal visits or admissions, number of hospital or health professional visits (including midwife, obstetrician, physician, dietician, diabetic nurse), length of antenatal stay, length of postnatal stay (maternal), length of postnatal stay (baby), cost of maternal care, cost of offspring care, costs associated with the intervention, costs to families associated with the management provided, cost of dietary monitoring (e.g. diet journals, dietician, nurse visits, etc), costs to families ‐ change of diet, extra antenatal visits, extra use of healthcare services (consultations, blood glucose monitoring, length and number of antenatal visits), women’s view of treatment advice, duration of stay in neonatal intensive care unit or special care baby unit).

Acarbose versus other oral anti‐diabetic agent (comparison 3)

We identified a single small study (n = 43) that compared glibenclamide with acarbose (Bertini 2005).

Maternal primary outcomes

3.1 Caesarean section

There was no evidence of a difference in the risk of caesarean section between women who had been treated with glibenclamide and those treated with acarbose (RR 0.95, 95% CI 0.53 to 1.70; one study, n = 43 women; low‐quality evidence; Analysis 3.1). The evidence was downgraded for risk of bias (method of randomisation was unclear) and imprecision (single small study).

3.1. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 1 Caesarean section.

Other maternal primary outcomes

No data were reported on hypertensive disorders of pregnancy or on the development of type 2 diabetes for this comparison.

Neonatal primary outcomes

3.2 Perinatal death and later infant mortality

There were no cases of perinatal death for either glibenclamide‐ or acarbose‐treated women reported by Bertini 2005. Absolute effects could not be reported due to no events being reported for the outcome. The quality of the evidence was low due to selective reporting and evidence being based on a single study.

3.3 LGA

Bertini 2005 found no evidence of a difference in the risk for being born LGA (greater than 90th percentile) between infants whose mothers had been treated with glibenclamide and those treated with acarbose (RR 2.38, 95% CI 0.54 to 10.46; one study, n = 43 women; Analysis 3.3). if the risk for being LGA was 10.5% in those infants whose mothers were treated with acarbose, between 5.7% to 100% of those whose mothers had been treated with glibenclamide would be born LGA. The quality of the evidence was judged to be low due to selective reporting and evidence being based on a single study.

3.3. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 3 Large‐for‐gestational age.

Other neonatal primary outcomes

No data were reported for death, serious morbidity composite or neurosensory disability in later childhood for this comparison.

Maternal secondary outcomes

3.4 Use of additional pharmacotherapy

Bertini 2005 reported no difference in the need for additional insulin between women who had been treated with glibenclamide and those treated with acarbose (RR 0.49, 95% CI 0.19 to 1.27; one study, n = 43 women; Analysis 3.4).

3.4. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 4 Need for additional pharmacotherapy.

3.5 Maternal hypoglycaemia

There were no events of maternal hypoglycaemia (requiring hospitalisation) for either glibenclamide‐ or acarbose‐treated women reported by Bertini 2005.

3.6 Weight gain in pregnancy

There was no evidence of a difference in weight gain in pregnancy between women who had been treated with glibenclamide and those treated with acarbose reported by Bertini 2005 (MD ‐0.60 Kg, 95% CI ‐3.13 to 1.93; one study, n = 43 women; Analysis 3.6).

3.6. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 6 Weight gain in pregnancy (Kg).

Other maternal secondary outcomes

No data were reported for the other maternal secondary outcomes of this review (glycaemic control during/end of treatment, adherence to the intervention, induction of labour, placental abruption, postpartum haemorrhage, postpartum infection, perineal trauma/tearing, breastfeeding at discharge, six weeks postpartum, six months or longer, maternal mortality, sense of well‐being and quality of life, behavioural changes associated with the intervention, views of the intervention, relevant biomarker changes associated with the intervention).

Long‐term maternal outcomes

No data were reported for long‐term maternal outcomes (postnatal depression, BMI, postnatal weight retention or return to pre‐pregnancy weight, type 1 diabetes, type 2 diabetes, impaired glucose tolerance, subsequent GDM, cardiovascular health (as defined by trialists including blood pressure, hypertension, cardiovascular disease, metabolic syndrome).

Neonatal secondary outcomes

3.7 Macrosomia

There was no evidence of a difference in the risk of macrosomia (greater than 4 kg) for infants whose mothers had been treated with glibenclamide and those treated with acarbose (RR 7.20, 95% CI 0.41 to 125.97; one study, n = 43 women; Analysis 3.7).

3.7. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 7 Macrosomia.

3.8 Small‐for‐gestational age

There were no events of being born small‐for‐gestational age (not defined) for either glibenclamide or acarbose treated women reported by Bertini 2005.

3.9 Birth trauma

There were no events of birth trauma (not specified) for either glibenclamide‐ or acarbose‐treated women reported by Bertini 2005.

3.10 Gestational age at birth

There was no evidence of a difference in gestational age at birth between infants whose mothers had been treated with glibenclamide and those treated with acarbose (MD ‐0.10 weeks, 95% CI ‐0.82 to 0.62; one study, n = 43 women; Analysis 3.10).

3.10. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 10 Gestational age at birth (weeks).

3.11 Preterm birth

There were no events of preterm birth for either glibenclamide‐ or acarbose‐treated women reported by Bertini 2005.

3.12 Birthweight

There was no evidence of a difference in birthweight between infants whose mothers had been treated with glibenclamide and those treated with acarbose (MD 153.00 g, 95% CI ‐123.52 to 429.52; one study, n = 43 women; Analysis 3.12 ).

3.12. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 12 Birthweight (Kg).

3.13 Neonatal hypoglycaemia

There was no evidence of a difference in the risk of neonatal hypoglycaemia (2.2 mmol/L; less than 40 mg/dL) for infants whose mothers had been treated with glibenclamide and those treated with acarbose (RR 6.33, 95% CI 0.87 to 46.32; one study, n = 43 women; low‐quality evidence; Analysis 3.13). If neonatal hypoglycaemia occurred in 5.3% of infants whose mothers had been treated with acarbose, the risk for those whose mothers had been treated with glibenclamide would range from 4.6% to 100%. The quality of the evidence was judged to be very low due to selective reporting, evidence being based on a single study and imprecision.

3.13. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 13 Neonatal hypoglycaemia.

3.14 Respiratory distress syndrome

There were no events of respiratory distress syndrome for either glibenclamide or acarbose treated women reported by Bertini 2005.

Other neonatal secondary outcomes

No other neonatal secondary outcomes were reported for this comparison (stillbirth, neonatal death, five‐minute Apgar less than seven, birthweight z score, head circumference and z score, length and z score, ponderal index, adiposity, neonatal jaundice (hyperbilirubinaemia), hypocalcaemia, polycythaemia, relevant biomarker changes associated with the intervention).

Later infant/childhood outcomes

No data were reported for childhood outcomes (weight and z scores, height and z scores, head circumference and z scores, adiposity, educational attainment, blood pressure, type 1 diabetes, type 2 diabetes, impaired glucose tolerance, dyslipidaemia or metabolic syndrome).

Child as an adult outcomes

No data were reported for child as an adult outcomes (weight, height, adiposity, cardiovascular health (as defined by trialists including blood pressure, hypertension, cardiovascular disease, metabolic syndrome), employment, education and social status/achievement, dyslipidaemia or metabolic syndrome, type 1 diabetes, type 2 diabetes, impaired glucose tolerance).

Health service outcomes

Admission to neonatal intensive care

One infant in the glibenclamide group was admitted to neonatal intensive care. The difference was not significant when compared with acarbose (RR 2.49, 95% CI 0.10 to 64.62; one study, n = 43).

Other health service outcomes

No data were reported for other health service outcomes (number of antenatal visits or admissions, number of hospital or health professional visits (including midwife, obstetrician, physician, dietician, diabetic nurse), admission to neonatal intensive care unit/nursery, length of antenatal stay, length of postnatal stay (maternal), length of postnatal stay (baby), cost of maternal care, cost of offspring care, costs associated with the intervention, costs to families associated with the management provided, cost of dietary monitoring (e.g. diet journals, dietician, nurse visits), costs to families ‐ change of diet, extra antenatal visits, extra use of healthcare services (consultations, blood glucose monitoring, length and number of antenatal visits), women’s view of treatment advice, duration of stay in neonatal intensive care unit or special care baby unit).

Discussion

Summary of main results

There was no evidence of a difference between women treated with oral anti‐diabetic pharmacological therapies compared with placebo for hypertensive disorders of pregnancy, induction of labour, perineal trauma or risk of birth by caesarean section (Table 1). Oral anti‐diabetic pharmacological therapy did appear to lower fasting capillary blood glucose concentrations compared with placebo in one study of 375 Hispanic women (Casey 2015). No data were reported for development of type 2 diabetes or return to pre‐pregnancy weight or postnatal depression. There was no evidence of a difference in the risk for being born LGA between infants whose mothers received glibenclamide or placebo (Table 2). There was no evidence of a difference in the risk of macrosomia or in birthweight between infants whose mothers had been treated with glibenclamide and those treated with placebo. No data were reported for perinatal mortality, death or serious morbidity composite, neonatal hypoglycaemia, diabetes or adiposity. Evidence was limited to two studies both of which used a different oral anti‐diabetic pharmacological therapy. Outcomes of interest for this review were poorly reported.

We found few differences between groups for the comparison of metformin versus glibenclamide. Metformin was associated with an increase in fasting blood glucose levels compared with glibenclamide, but no difference in postprandial glucose levels (Table 3). There was no evidence of a difference between groups for hypertensive disorders of pregnancy, birth by caesarean section, perineal trauma or induction of labour. No data were reported for postnatal depression or development of type 2 diabetes. Metformin was associated with a reduced risk of death or serious morbidity composite (hypoglycaemia, hyperbilirubinaemia, macrosomia, respiratory illness, birth injury, stillbirth or neonatal death) compared with glibenclamide. There was no evidence of a difference between groups for being born LGA, perinatal mortality or neonatal hypoglycaemia. No data were reported for adiposity of diabetes (Table 4). The evidence for each outcome was based on a limited number of studies with low sample size.

The limited evidence for glibenclamide versus acarbose did not identify any differences between groups for the limited maternal and infant outcomes that were reported. There was no evidence of a difference between groups for birth by caesarean section. No data were reported for hypertensive disorders of pregnancy, induction of labour, development of type 2 diabetes, perineal trauma, return to pre‐pregnancy weight or postnatal depression (Table 5). There were no events reported in either group for perinatal mortality. There was no evidence of a difference between groups for being born LGA or neonatal hypoglycaemia. No data were reported for death or serious morbidity composite, adiposity or diabetes (Table 6).

Overall completeness and applicability of evidence

All of the studies reported on women with GDM, although the proportion of women with GDM in the Notelovitz 1971 study could not be clearly ascertained and therefore we could not include the data in a meta‐analysis. The number of studies identified was limited for some of the comparisons. In one study we questioned the generalisability of the data due to the demographics of the women recruited being 93% Hispanic (Casey 2015). Oral anti‐diabetic agent versus placebo or usual care was reported as a comparison by three studies(Cortez 2006; Myers 2014; Notelovitz 1971); metformin versus glibenclamide was reported by four studies (Fenn 2015; George 2015; Moore 2010; Silva 2012); one study compared glibenclamide with acarbose (Bertini 2005) and one study compared glibenclamide with the addition of metformin if glycaemic targets were not met against metformin with the addition of glibenclamide if glycaemic targets were not met (Nachum 2015).

No data were reported on long‐term maternal or neonatal/child outcomes and no data were reported on direct or indirect costs of the interventions. Health service outcomes were very poorly reported.

Clinicians and women may also be interested in the comparison of oral anti‐diabetic pharmacological therapies compared with insulin and we refer the reader to another Cochrane Review that includes this comparison (Brown 2016).

None of the studies reported long‐term GRADE outcomes for development of type 2 diabetes and postnatal depression for the mother, and none reported adiposity and diabetes for the infants.

Quality of the evidence

Eleven included studies (1487 women and their babies) were identified. Two studies are awaiting classification. Overall we judged the evidence to be of unclear risk of bias due to lack of methodological details provided in the individual studies.

Two studies reported data for the comparison of oral anti‐diabetic pharmacological therapies versus placebo, the main methodological limitations were a lack of adequate reporting to make a judgement regarding risk of bias, lack of generalisability and evidence being based on a single study for many of the reported outcomes. The overall quality of the evidence was judged to be very low using GRADE methodology (Table 1; Table 2.

Six studies reported data for the comparison of metformin and glibenclamide, the most commonly used oral anti‐diabetic pharmacological therapies. The evidence was generally of moderate to low quality and was downgraded due to lack of blinding, imprecision and evidence being based on a single trial (Table 3; Table 4).

One study reported on the comparison of glibenclamide versus acarbose, the evidence was considered to be of low quality in the one outcome of interest for this review that was reported and used for assessment of quality using GRADE. The evidence was downgraded for unclear randomisation and only being based on a single study (Table 5; Table 6).

Potential biases in the review process

We conducted a comprehensive search of the literature including both full papers and conference abstracts with no language restrictions. Two researchers independently undertook identification of studies for inclusion in this review, data extraction and data entry. The main limitation of this systematic review is the lack of studies overall and a lack of studies for each comparison. We could not include any data in meta‐analyses for chlorpropamide or tolbutamide. These drugs are known to cross the placenta and, due to concerns about risks of congenital abnormalities, are not recommended during pregnancy in high‐ and middle‐income countries. We could not estimate the extent of their use in low‐income countries. There was a lack of data reported for the outcomes of interest for this review. Long‐term outcomes were not reported.

Agreements and disagreements with other studies or reviews

A systematic review by Balsells 2015 included a head‐to‐head comparison of metformin with glibenclamide and included two studies (Moore 2010; Silva 2012), both of which are included in this review. Our results concur with the majority of their findings although Balsells 2015 reported that metformin was associated with a reduction in the risk of being born LGA or macrosomic whereas our analyses suggest no difference between metformin and glibenclamide for these outcomes. Another systematic review (Amin 2015) compared glibenclamide with metformin in three included studies (George 2015; Moore 2010; Silva 2012). Their data and that of our review are in agreement, with the exception of LGA and macrosomia, which the Amin 2015 systematic review combined in a composite outcome that reflected an increased risk in infants whose mothers had been treated with glibenclamide (RR 1.94, 95%CI 1.03 to 3.66; three studies, 508 infants) and for which our systematic review reports data separately, finding no evidence of a difference either for macrosomia (RR 0.72, 95%CI 0.23 to 2.21; two studies, 308 infants) or LGA (RR 0.67, 95%CI 0.24 to 1.83; two studies, 246 infants).

Authors' conclusions

Implications for practice.

Due to lack of data reported for the outcomes of interest in this review, the evidence is uncertain as to whether one oral anti‐diabetic pharmacological therapy is safer or more effective than another. The choice in use of oral anti‐diabetic pharmacological therapy is likely dependent on factors such as preference of the woman, availability and national clinical practice guidelines.

Implications for research.

Oral anti‐diabetic pharmacological therapies are becoming more widely used for treating women with gestational diabetes mellitus (GDM). Uncertainty remains due to the limited evidence to support the use of one oral anti‐diabetic pharmacological therapy over another.

Future research trials should be encouraged to report on the core outcomes suggested in this review and in particular the long‐term outcomes for the woman and the infant that have been poorly reported to date.

Notes

The original review, Alwan 2009 has been split into three new reviews due to the complexity of the included interventions. The following new reviews are underway.

• Lifestyle interventions for the treatment of women with gestational diabetes mellitus (Brown 2015)

• Oral anti‐diabetic pharmacological therapies for the treatment of women with gestational diabetes mellitus (this review)

• Insulin for the treatment of women with gestational diabetes mellitus (Brown 2016)

Appendices

Appendix 1. Trial registry search strategy

ClinicalTrials.gov and WHO ICTRP

Search terms: oral anti‐diabetic OR oral hypoglycaemic* OR oral hypoglycemic* OR metformin OR glibenclamide OR glyburide OR acarbose AND gestational diabetes OR GDM

(In ICTRP, each term for medication was combined separately with each term for gestational diabetes)

Data and analyses

Comparison 1. Oral anti‐diabetic agents versus placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Hypertensive disorders of pregnancy	1		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
1.1 Hypertensive disorders of pregnancy (any type)	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.24 [0.81, 1.90]
1.2 Pregnancy‐induced hypertension	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.24 [0.71, 2.19]
1.3 Pre‐eclampsia	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.23 [0.59, 2.56]
2 Caesarean section	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.03 [0.79, 1.34]
3 Large‐for‐gestational age	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.51, 1.58]
4 Use of additional pharmacotherapy	2	434	Risk Ratio (M‐H, Fixed, 95% CI)	0.68 [0.42, 1.11]
4.1 Placebo	2	434	Risk Ratio (M‐H, Fixed, 95% CI)	0.68 [0.42, 1.11]
5 Glycaemic control (end of treatment) (mg/dL)	1	375	Mean Difference (IV, Fixed, 95% CI)	‐3.0 [‐5.13, ‐0.87]
6 Weight gain in pregnancy (Kg)	1	375	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.96, 0.96]
7 Induction of labour	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.18 [0.79, 1.76]
8 Perineal trauma	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.06, 15.62]
9 Stillbirth	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	0.49 [0.05, 5.38]
10 Neonatal death	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
11 Small‐for‐gestational age	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.11 [0.58, 2.10]
12 Macrosomia	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	0.71 [0.36, 1.41]
13 Birthweight (g)	1	375	Mean Difference (IV, Fixed, 95% CI)	‐33.0 [‐134.53, 68.53]
14 Shoulder dystocia	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	0.33 [0.01, 8.00]
15 Bone fracture	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	0.74 [0.17, 3.25]
16 Nerve palsy	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	0.33 [0.01, 8.00]
17 Gestational age at birth (weeks)	1	375	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐0.32, 0.32]
18 Neonatal hypoglycaemia	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.97 [0.36, 10.62]
19 Hyperbilirubinaemia	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.97 [0.50, 7.75]
20 Admission to NICU	1	375	Risk Ratio (M‐H, Fixed, 95% CI)	1.16 [0.53, 2.53]

1.10. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 10 Neonatal death.

1.13. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 13 Birthweight (g).

1.16. Analysis.

Comparison 1 Oral anti‐diabetic agents versus placebo, Outcome 16 Nerve palsy.

Comparison 2. Metformin versus glibenclamide.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Hypertensive disorders of pregnancy	3	508	Risk Ratio (M‐H, Fixed, 95% CI)	0.70 [0.38, 1.30]
1.1 Pre‐eclampsia	1	149	Risk Ratio (M‐H, Fixed, 95% CI)	0.66 [0.11, 3.82]
1.2 Pregnancy‐induced hypertension	2	359	Risk Ratio (M‐H, Fixed, 95% CI)	0.71 [0.37, 1.37]
2 Caesarean section	4	554	Risk Ratio (M‐H, Random, 95% CI)	1.20 [0.83, 1.72]
2.1 Carpenter and Coustan criteria	2	195	Risk Ratio (M‐H, Random, 95% CI)	2.36 [0.53, 10.52]
2.2 National Diabetes Data Group criteria	1	159	Risk Ratio (M‐H, Random, 95% CI)	1.12 [0.75, 1.68]
2.3 World Health Organization (1999)	1	200	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.78, 1.15]
3 Perinatal mortality	2	359	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.06, 14.55]
4 Large‐for‐gestational age	2	246	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.24, 1.83]
4.1 Carpenter and Coustan criteria	1	46	Risk Ratio (M‐H, Random, 95% CI)	1.25 [0.38, 4.07]
4.2 World Health Organization (1999)	1	200	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.21, 0.92]
5 Death or serious morbidity composite	1	159	Risk Ratio (M‐H, Fixed, 95% CI)	0.54 [0.31, 0.94]
6 Use of additional pharmacotherapy	5	660	Risk Ratio (M‐H, Random, 95% CI)	0.66 [0.28, 1.57]
7 Maternal hypoglycaemia	3	354	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.36, 2.19]
8 Glycaemic control (mg/L; mmol/L)	3		Std. Mean Difference (IV, Fixed, 95% CI)	Subtotals only
8.1 Fasting blood glucose	3	508	Std. Mean Difference (IV, Fixed, 95% CI)	0.19 [0.02, 0.37]
8.2 Postprandial blood glucose	3	508	Std. Mean Difference (IV, Fixed, 95% CI)	0.16 [‐0.01, 0.34]
8.3 HbA1c	1	200	Std. Mean Difference (IV, Fixed, 95% CI)	‐0.12 [‐0.39, 0.16]
9 Weight gain in pregnancy (Kg)	1	200	Mean Difference (IV, Fixed, 95% CI)	‐2.06 [‐3.98, ‐0.14]
10 Induction of labour	1	159	Risk Ratio (M‐H, Fixed, 95% CI)	0.81 [0.61, 1.07]
11 Perineal trauma	2	308	Risk Ratio (M‐H, Fixed, 95% CI)	1.67 [0.22, 12.52]
12 Stillbirth	1	200	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.06, 14.55]
13 Macrosomia	2	308	Risk Ratio (M‐H, Fixed, 95% CI)	0.72 [0.23, 2.21]
14 Birth trauma	1	159	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
15 Shoulder dystocia	2	195	Risk Ratio (M‐H, Fixed, 95% CI)	0.99 [0.14, 6.89]
16 Gestational age at birth (weeks)	3	508	Mean Difference (IV, Fixed, 95% CI)	0.03 [‐0.22, 0.28]
17 Preterm birth	3	508	Risk Ratio (M‐H, Fixed, 95% CI)	1.59 [0.59, 4.29]
18 5‐minute Apgar < 7	1	149	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
19 Birthweight (g)	2	349	Mean Difference (IV, Fixed, 95% CI)	‐209.13 [‐314.53, ‐103.73]
20 Ponderal index	1	200	Mean Difference (IV, Fixed, 95% CI)	‐0.09 [‐0.17, ‐0.01]
21 Neonatal hypoglycaemia	4	554	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.42, 1.77]
22 Respiratory distress syndrome	1	159	Risk Ratio (M‐H, Fixed, 95% CI)	0.51 [0.10, 2.69]
23 Hyperbilirubinaemia	2	205	Risk Ratio (M‐H, Fixed, 95% CI)	0.68 [0.37, 1.25]
24 Admission to NICU	2	349	Risk Ratio (M‐H, Fixed, 95% CI)	1.52 [0.65, 3.56]

2.14. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 14 Birth trauma.

2.18. Analysis.

Comparison 2 Metformin versus glibenclamide, Outcome 18 5‐minute Apgar < 7.

Comparison 3. Glibenclamide versus acarbose.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Caesarean section	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	0.95 [0.53, 1.70]
2 Perinatal mortality	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
3 Large‐for‐gestational age	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	2.38 [0.54, 10.46]
4 Need for additional pharmacotherapy	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	0.49 [0.19, 1.27]
5 Maternal hypoglycaemia	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
6 Weight gain in pregnancy (Kg)	1	43	Mean Difference (IV, Fixed, 95% CI)	‐0.60 [‐3.13, 1.93]
7 Macrosomia	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	7.20 [0.41, 125.97]
8 Small‐for‐gestational age	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
9 Birth trauma (not specified)	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
10 Gestational age at birth (weeks)	1	43	Mean Difference (IV, Fixed, 95% CI)	‐0.10 [‐0.82, 0.62]
11 Preterm birth	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
12 Birthweight (Kg)	1	43	Mean Difference (IV, Fixed, 95% CI)	153.0 [‐123.52, 429.52]
13 Neonatal hypoglycaemia	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	6.33 [0.87, 46.32]
14 Respiratory distress syndrome	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]

3.2. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 2 Perinatal mortality.

3.5. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 5 Maternal hypoglycaemia.

3.8. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 8 Small‐for‐gestational age.

3.9. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 9 Birth trauma (not specified).

3.11. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 11 Preterm birth.

3.14. Analysis.

Comparison 3 Glibenclamide versus acarbose, Outcome 14 Respiratory distress syndrome.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Bertini 2005.

Methods	Randomised, parallel, open‐label study
Participants	70 women Inclusion criteria: women diagnosed with GDM whose glycaemic targets were not adequately controlled by diet and exercise alone. GDM diagnosed using WHO criteria 75 g OGTT fasting ≥ 6.1 mmol/L (110 mg/dL); 2‐h value ≥ 7.8 mmol/L (140 mg/dL). Gestational age 11‐33 weeks, singleton pregnancy Exclusion criteria: presence of a pathology requiring faster glucose control (e.g. antenatal corticosteroids), other pathologies affecting therapy or perinatal results (no details) Setting: maternity hospital Joinville SC, Brazil Timing: October 2003‐July 2004
Interventions	Glyburide (n = 24) ‐ initial dose 5 mg in the morning, increasing every 7 d until glycaemic control achieved up to a maximum of 20 mg Acarbose (n = 19) ‐ initial dose 50 mg before main meals with 50 mg increments every 7 d until glycaemic control achieved to a maximum of 300 mg Insulin (n = 27) ‐ not applicable for this systematic review Where maximum dose was met without adequate glycaemic control insulin therapy was commenced
Outcomes	No primary outcomes were listed for the mother. Secondary outcomes included fasting and postprandial glucose levels, gestational age at birth, severe hypoglycaemia requiring hospitalisation, BMI, gestational weight gain, type of delivery, other occurrences For the infant, primary outcomes: fetal weight, fetal hypoglycaemia; secondary outcomes: birthweight, macrosomia, LGA, capillary blood glucose, neonatal hypoglycaemia, bilirubin level, calcium level, duration of hospitalisation, admission to NICU, death, discharge status
Notes	Sample size calculation ‐ no details ITT analysis ‐ no Funding ‐ no details
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	State that randomised but no details
Allocation concealment (selection bias)	Low risk	"Randomization was done by using brown envelopes containing outside the randomization number and in the inside a sheet defining which therapy the patient was allocated to"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Open label study. "SInce this study compares 3 therapies with different administration procedures, this was not a blind study"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Open label study. No details as to whether outcome assessors were blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	71 women in total were randomised and 1 was later excluded due to severe asthma that required corticotherapy. This woman was excluded from further analysis
Selective reporting (reporting bias)	Low risk	No evidence of selective reporting although an original protocol was not viewed
Other bias	Low risk	No differences in baseline

Casey 2015.

Methods	Randomised, parallel study
Participants	395 women Inclusion criteria: at least 2 abnormal values on a 3‐h 100 g OGTT using NDDG criteria and fasting values > 5.8 mmol/L (105 mg/dL); 24‐30 weeks' gestation; singleton pregnancy Exclusion criteria: established pre‐gestational diabetes; abnormal gestational diabetes screening (≥ 140 mg/dL) prior to 24 weeks' gestation, multiple pregnancy; known major fetal anomaly or fetal demise; any renal disease with serum creatinine > 1.0 mg/dL; known liver disease such as hepatitis; maternal or fetal conditions likely to require imminent or very preterm delivery such as pre‐eclampsia, preterm premature rupture of membranes, preterm labour, and IUGR; known hypersensitivity or allergic reaction to glyburide Setting: medical centre, Dallas, Texas, USA Timing: September 2008‐October 2012
Interventions	All women underwent monitored diet with weekly diary logs and 4 times daily glucose monitoring. Treatment targets were fasting < 5.3 mmol/L (95 mg/dL) and < 6.7 mmol/L (120 mg/dL) for 2‐h post‐prandial glucose readings Glibenclamide (n = 189) starting dose 2.5 mg and titrated up to a maximum of 20 mg/d based on weekly maternal capillary glucose readings Placebo (n = 186) identical capsule to glibenclamide
Outcomes	Primary ‐ birthweight decrease of 200 g Secondary outcomes ‐ mean capillary blood glucose, need for insulin, chorioamnionitis, pregnancy‐induced hypertension, need for operative birth, shoulder dystocia, perineal trauma, maternal weight gain, birthweight, SGA, LGA, admission to neonatal intensive care, fracture clavicle, Erbs palsy, hyperbilirubinaemia, active treatment of hypoglycaemia, cord blood pH </= 7
Notes	Trial registration NCT00744965 Sample size calculation ‐ yes based on birthweight Funding ‐ Department of Obstetrics and Gynaecology ITT analysis ‐ no
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated randomization schedule"
Allocation concealment (selection bias)	Low risk	Masking, allocation and assignment done by the investigational drug pharmacy
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants blinded with identical placebo capsule
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Outcome assessors were blinded according to trial registration
Incomplete outcome data (attrition bias) All outcomes	Low risk	20 women lost to follow‐up: glibenclamide n = 9 did not give birth in study hospital or lost to follow‐up; placebo n = 11 did not give birth in study hospital or lost to follow‐up
Selective reporting (reporting bias)	High risk	No published protocol found but the trial registration listed primary and secondary outcomes. However, many more maternal and neonatal outcomes were reported in the published paper than were prespecified in the trial registration document
Other bias	High risk	This trial appears to have been registered twice as NCT00942552 and as NCT 00744965 with the same outcomes, interventions and sample size. The population was 93% Hispanic and therefore the results may not be generalisable to other ethnicities

Cortez 2006.

Methods	Randomised, parallel study
Participants	59 women Women diagnosed with gestational diabetes (no details for diagnosis provided) between 12 and 34 weeks Conference abstract with no details of inclusion or exclusion criteria Setting: Callifornia, USA Timing: not specified
Interventions	Acarbose 50 mg 3 times/d and increased by 50 mg if 50% of glycaemic targets readings were not achieved in a week (fasting < 5.3 mmol/L, < 95 mg/dL; 1‐hour postprandial < 7.5 mmol/L, < 135 mg/dL) (n = 29) Placebo (n = 30) no details
Outcomes	Treatment failure, weight gain, side effects, gestational age at delivery, Apgar scores, birthweight, mode of delivery
Notes	Conference abstract only. Unable to find contact details for study authors, no evidence of a full paper publication and no definition of GDM provided ITT analysis ‐ states it has used ITT analysis Power calculation ‐ yes based on failure rates Funding ‐ no details
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	States randomised but no details
Allocation concealment (selection bias)	Unclear risk	No details
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	States double blind but no details
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	States double blind but no details
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No details provided, although state that ITT analysis conducted
Selective reporting (reporting bias)	High risk	Conference abstract only published in 2006, no full publication identified. Not all data reported
Other bias	High risk	Unable to judge as conference proceeding only

De Bacco 2015.

Methods	Randomised, parallel study
Participants	81 women Inclusion criteria: "eligible women between 18‐45 years diagnosed with gestational diabetes, according to the WHO criteria, ratified the public network for prenatal care at the Clinic for Diabetes and Pregnancy. Women with singleton pregnancy and gestational age not exceeding 30 weeks at the time of enrolment, calculated from ultrasound (U.S.) Obstetric performed before the 20th week of pregnancy will be enrolled" (taken from clinical trial registry NCT02091336) Exclusion criteria: "women who present enrolment in the diagnosis of chronic hypertension, heart disease or chronic lung disease intrauterine restricted or preterm labour, growth, or even chronic diarrhoea will be excluded" (taken from clinical trial registry NCT02091336) Setting: Hosptial de Clinicas, Porto Allegre, Brazil Timing: not specified
Interventions	Glibenclamide (n = 45). Initial dose of 2.5 mg/d, increased by 2.5 mg the following week, and following increments of 5 mg/week until glycaemic control achieved or maximum dose of 20 mg/d Metformin (n = 36). Initial dose of 500 mg, increasing in 500 mg increments every 3 d until glycaemic target achieved or maximum dose of 2.5 g/d
Outcomes	Caesarean section, perinatal death, LGA, need for additional pharmacological therapy, maternal hypoglycaemia, weight gain in pregnancy, macrosomia, SGA, birth trauma (any), gestational age at birth, preterm birth (< 37 weeks), birthweight, neonatal hypoglycaemia, RDS
Notes	Additional information was obtained from the clinical trial registration NCT02091336 Data only currently reported in abstract format
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Randomized" no further details
Allocation concealment (selection bias)	Unclear risk	No details
Blinding of participants and personnel (performance bias) All outcomes	High risk	Open label study
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	High risk	8/36 women dropped out of metformin arm (hypoglycaemia n = 1; lack of glycaemic control n = 1; gastric intolerance n = 6) and 24/45 dropped out of glibenclamide arm (hypoglycaemia n = 17; lack of glycaemic control n = 6; gastric intolerance n = 1)
Selective reporting (reporting bias)	High risk	In this abstract the only data reported were for dropouts and not for those who remained in the study
Other bias	High risk	Data reported as conference abstract only. No baseline data were presented

Fenn 2015.

Methods	Randomised, parallel study, single centre
Participants	48 women Inclusion criteria: diagnosed with GDM using Carpenter and Coustan criteria, failed to meet treatment targets with diet alone after 2 weeks Exclusion criteria: not stated Setting: Kerala, India Timing: not stated
Interventions	Glibenclamide (n = 24) initial dose 2.5 mg twice daily increasing to a maximum of 10 mg daily. If treatment targets not met then insulin added Metformin (n = 24) initial dose 500 mg twice daily, increasing to a maximum dose of 1700 mg daily. If treatment targets not met then insulin added
Outcomes	Maternal glycaemic control (HbA1c in third trimester), birthweight, cord C peptide. Mode of delivery, shoulder dystocia, neonatal blood sugars at 3 and 6 h, hyperbilirubinaemia, incidence of urinary tract infection, incidence of candidiasis, pre‐eclampsia, weight gain (10 kg or more)
Notes	Sample size calculation ‐ no details ITT analysis ‐ no Funding ‐ no details provided Conflicts of interest ‐ authors state that there were no financial or other conflicts of interest associated with this study
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block randomisation using computer‐generated random numbers
Allocation concealment (selection bias)	Unclear risk	No details
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Physician and patient were blinded to the allocation. Tablets dispensed in sealed envelopes from the pharmacy
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 woman in each arm was lost to follow‐up. Analysed 46 women
Selective reporting (reporting bias)	High risk	Some additional outcomes were reported that were not specified in the methods section
Other bias	Unclear risk	Unable to ascertain if demographics similar at baseline as no data provided

George 2015.

Methods	Randomised, parallel‐controlled study, single centre
Participants	159 women Mean age of glibenclamide group was 33.6 ± 4.6 years and for the metformin group was 33.4 ± 4.4 years. Mean BMI glibenclamide group 28.8 ± 4.0 kg/m2 and for the metformin group was 28.7 ± 4.4 kg/m2 All women screened by risk factors followed by 100 g OGTT between 24‐28 weeks' gestation. Criteria for diagnosis was NDDG (1979) using 2 abnormal values from fasting glucose ≥ 5.3 mmol/L, 1‐h ≥ 10 mmol/L, 2‐h ≥ 8.6 mmol/L, 3‐h ≥ 7.8 mmol/L. Gestational age at recruitment was 29.7 ± 3.7 weeks in the glibenclamide group and 29.3 ± 3.3 weeks in the metformin group Inclusion criteria: failed to meet treatment targets using medical nutritional therapy (fasting glucose ≥ 5.5 mmol/L and ≤ 7.8 mmol/L, 2‐h postprandial ≥ 6.7 mmol/L and ≤ 13.9 mmol/L), 20‐33 weeks' gestation Exclusion criteria: pre‐existing type 1 or type 2 diabetes, currently taking metformin for some other indication, multiple pregnancy, recognised fetal anomaly, known abnormal renal or hepatic function, hypoxic cardio‐respiratory disease, malabsorption or some other significant gastrointestinal disease, sepsis, ruptured membranes, gestational hypertension or pre‐eclampsia Setting: India Timing: 2007‐2010
Interventions	Glibenclamide (n = 80) initial dose 2.5 mg. Doses increased weekly if required to a maximum of 15 mg/d Metformin (n = 79) initial dose 500 mg/d, increased weekly if required to a maximum of 2500 mg/d Blood sugars self‐monitored 4 times/week Treatment target during therapy fasting glucose ≤ 5.3 mmol/L and 2‐h postprandial ≤ 6.7 mmol/L. If targets were not met in 2‐3 weeks insulin was added or women were switched over to insulin alone All women were induced not later than 39 weeks' gestation
Outcomes	Primary outcomes: composite including macrosomia (> 3.7 kg), hypoglycaemia (≤ 2.2 mmol/L), need for phototherapy, RDS, stillbirth, neonatal death, birth trauma Secondary outcomes: birthweight, maternal glycaemic control, hypertension, preterm birth < 34 weeks, induction of labour, mode of birth, complications of birth
Notes	Power calculation: yes based on composite outcome ITT analysis: yes Funding: none specified Conflicts of interest: not detailed in manuscript
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated random list"
Allocation concealment (selection bias)	Low risk	"sequentially labelled opaque envelopes" "arranged...in a central research office by research officers not involved in patient care"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Women were not blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Research officers collecting data were masked to allocation. After birth all babies were monitored by neonatologists who were masked to study participation
Incomplete outcome data (attrition bias) All outcomes	Low risk	In the glibenclamide group 6 women did not receive the allocated intervention (obstetrician withdrew 1 woman, 4 women withdrew, 1 woman gave birth elsewhere) In the metformin group 4 women did not receive the allocated intervention (1 withdrew from study and 3 gave birth elsewhere)
Selective reporting (reporting bias)	Low risk	The outcome of preterm birth < 34 weeks' gestation was not listed as an outcome in the Trial Registration on the Clinical Trials Registry India (CTRI/2014/02/004418 ‐ registered retrospectively) but was listed and reported in the published manuscript
Other bias	Unclear risk	An interim analysis requested by the local data monitoring committee showed significant differences in outcomes and the study was stopped before the total sample size of 86 women per group was achieved Groups were balanced at baseline although the metformin group had higher fasting triglyceride levels

Moore 2010.

Methods	Randomised, parallel study
Participants	149 women diagnosed with GDM following a 1‐h 50 g OGCT (≥ 7.2 mmol/L; 130 mg/dL) followed by a 3‐h 100 g OGTT using Carpenter and Coustan criteria with 2 or more abnormal results. All women initially treated with diet and exercise counselling. Treatment targets were fasting 5.8 mmol/L (105 mg/dL); or 2‐h postprandial blood glucose level 6.7 mmol/L (120 mg/dL). Mean age of women in the glyburide group was 29.6 ± 7.8 years and in the metformin group was 31 ± 7.1 years. Mean BMI in glyburide group was 32.7 ± 7.0 kg/m2 and in the metformin group was 32.8 ± 5.8 kg/m2. Both groups comprised 88% Hispanic women Inclusion criteria: between 11 and 33 weeks' gestation Exclusion criteria: history of significant renal or hepatic disease, chronic hypertension requiring medication, or substance misuse Setting: University of New Mexico, Albuquerque, USA Timing: July 2003‐May 2008.
Interventions	Metformin (n = 75) initial dose of 500 mg per day taken in divided doses and increased as required to a maximum of 2000 g/d Glibenclamide (n = 74) initial dose of 2.5 mg twice daily increased as required to a maximum of 20 mg daily
Outcomes	Primary outcome: glycaemic control Secondary outcomes: medication failure rate, macrosomia (> 4000 g), admission to NICU, 5‐minute Apgar less than 7, birth trauma, pre‐eclampsia, maternal and neonatal hypoglycaemia, route of delivery
Notes	Elective delivery was planned at 38 weeks by induction of labour or repeat caesarean section as required ITT ‐ all randomised women were included in the analysis Power calculation ‐ yes, based on a difference in glycaemic control between groups Funding ‐ no details Conflicts of interest ‐ authors state in the manuscript that there were no financial or other conflicts
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"computer generated random list"
Allocation concealment (selection bias)	Low risk	"Sequentially labelled, opaque, sealed envelopes"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Study participants and care providers were not blinded to the treatment allocation
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	All women randomised were analysed. In the glyburide group 3 women never took the drug and 3 relocated before birth. In the metformin group, 5 women had only 2 prenatal visits, 2 relocated and 1 could not tolerate the gastrointestinal side effects and only took 2 doses of metformin
Selective reporting (reporting bias)	Low risk	All outcomes prespecified in the methods were reported in the results section
Other bias	Low risk	There were no group differences at baseline

Myers 2014.

Methods	Parallel, randomised controlled trial (pilot study)
Participants	40 women (target for study was 60) Inclusion criteria: women with mild GDM (fasting blood glucose 5.1‐5.4 mmol/L, 2‐h < 8.5 mmol/L) Exclusion criteria: multiple pregnancy, previous stillbirth, previous shoulder dystocia requiring obstetric manoeuvres, < 16 years old, unable to consent, known allergy or contra‐indication to study medication, liver abnormalities, renal dysfunction, acute or chronic disease which might cause tissue hypoxia, lactation Setting: Manchester, UK Timing: unclear
Interventions	Metformin up to 2000 mg/daily with initial dose of 500 mg/day from 26‐28 weeks' gestation to birth +/‐ insulin if treatment targets not met (no home monitoring) (n = 18) Standard care with dietary advice +/‐ metformin or insulin if treatment targets not met. Self‐monitoring of blood glucose pre‐meal and 1‐h postprandial (n = 19)
Outcomes	Anxiety, blood glucose, serum insulin, HOMA‐IR, acceptability, need for insulin (taken from trial registration document)
Notes	EudraCT number: 2013‐004065‐13/ ISRCTN86503951
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Low risk	Trial registration document states "..prefilled sealed envelopes created by independent research midwives within the department"
Blinding of participants and personnel (performance bias) All outcomes	High risk	"open label"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	3 women did not complete the trial and were not analysed, unclear as to which group they were allocated
Selective reporting (reporting bias)	High risk	No prespecified outcomes were provided in the conference abstract. Data were reported on n = 40 when according to the trial registration documentation the sample size was n = 60
Other bias	High risk	No reported differences at baseline although differences were reported between participants and non‐participants. Evidence in conference abstract format only at present

Nachum 2015.

Methods	Randomised, parallel controlled study
Participants	106 women Inclusion criteria: women diagnosed with GDM using Carpenter and Coustan criteria, 14‐33 weeks' gestation, aged 18‐45 years, 1 week of dietary treatment, sonographic dating of the pregnancy earlier than 24 weeks Exclusion criteria: suspected IUGR earlier than 24 weeks' gestation, major fetal malformation, pregestational diabetes mellitus Setting: Israel Timing: 2012‐2014
Interventions	Glibenclamide (n = 55) maximum dose 20 mg per day Metformin (n = 51) maximum dose 2550 mg per day If treatment targets not met then the other drug was added. If both failed then insulin was added Self‐monitoring of blood sugar 7 times/d
Outcomes	Primary outcome ‐ glycaemic control
Notes	Sample size calculation ‐ yes based on glycaemic control ITT analysis ‐ not clear in conference abstract Funding ‐ no details in conference abstract Conflicts of interest ‐ no details of whether there was a conflict was stated in the conference abstract
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	States randomised but no details
Allocation concealment (selection bias)	Unclear risk	No details
Blinding of participants and personnel (performance bias) All outcomes	High risk	Open label, participants were not blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	High risk	Unclear as the data were only reported in a conference abstract
Selective reporting (reporting bias)	High risk	Unclear as outcomes appear to be reported that were not listed in methods section or in the study registration document
Other bias	High risk	Data were presented as a conference abstract only

Notelovitz 1971.

Methods	Randomised parallel trial
Participants	207 women from South Africa Inclusion criteria: women who had been screened using a 2‐h 100 g OGTT with blood glucose values ≥ 7.8 mmol/L (140 mg/dL), remaining duration of pregnancy allowing for 6 weeks of intervention. Included women with known diabetes, glycosuria, family and obstetric histories suggestive of diabetes Exclusion criteria: established diabetics already on a specific treatment were not randomised Setting: Durban, South Africa Timing: not stated
Interventions	Chlorpropramide (n = 58) maximum dose 250 mg/day Tolbutamide (n = 46) maximum dose 1.5 g/day Insulin (n = 47) Dietary restriction alone (n = 56) Those participants who failed to respond to a treatment were usually then given insulin
Outcomes	None prespecified
Notes	Power calculation ‐ not stated ITT analysis ‐ yes Funding ‐ financial support received from Pfizer laboratories Conflicts of interest ‐ not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"random sample basis", no other details
Allocation concealment (selection bias)	Unclear risk	No details
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	Low risk	All women randomised appear to have data
Selective reporting (reporting bias)	High risk	There were no pre‐specified outcomes for the mother or the infant
Other bias	High risk	Paper states that there were no differences between interventions at baseline. The data for GDM and other diabetes could not be separated. The proportion of women with GDM could not be determined and the data have therefore not been included in any meta‐analysis

Silva 2012.

Methods	Randomised, parallel study
Participants	200 women. Mean age in metformin group 32.6 ± 5.6 years and for glyburide group 31.3 ± 5.4 years Women with GDM requiring additional pharmacotherapy. Diagnosis was by WHO criteria Inclusion criteria: > 18 years of age, gestational age 11‐33 weeks, single gestation, fetal abdominal circumference > 10% and < 75%, absence of other pathologies that might interfere with perinatal results or hypoglycaemic therapy. Capilliary glucose testing fasting 5.0 mmol/L (90 mg/dL), 1‐h postprandial after breakfast lunch and supper < 6.7 mmol/L (120 mg/dL); 2 abnormal values required Exclusion criteria: intolerance to drugs, unwillingness to participate, fetal risk (abdominal circumference < 10% or > 97%), lack of follow‐up or fetal malformation diagnosed at birth Setting: hospital medical centre, Santa Catarina, Brazil Timing: July 2008‐September 2010
Interventions	Glyburide (n = 96) starting dose 2.5 mg before breakfast and dinner and increased by 2.5 to 5 mg/week until glycaemic control achieved or to a maximum of 20 mg/d Metformin (n = 104) starting dose 500 mg at breakfast and dinner and increased by 500‐1000 mg weekly until glycaemic control achieved or a maximum of 2500 mg/d Insulin was started at 0.7 IU/kg/day regular insulin preprandial and NHP insulin at bedtime when glycaemic targets were not met
Outcomes	Primary outcomes ‐ maternal glucose control, weight, neonatal glucose levels. Maternal ‐ weight gain during pregnancy, need for change in therapy, HbA1c, ketonuria, gestational age at birth, severe hypoglycaemia requiring hospitalisation, mode of birth, complications with hypertensive disorders Neonatal ‐ birthweight, LGA, macrosomia, fetal hypoglycaemia, CGT after birth, duration of hospitalisation, death, hospital discharge conditions
Notes	Sample size calculations: no details Funding: no details ITT analysis: where possible
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"randomized", "computer generated randomization"
Allocation concealment (selection bias)	Low risk	Sequential numbering in brown envelopes with the name of the group glyburide or metformin
Blinding of participants and personnel (performance bias) All outcomes	High risk	"Open clinical study"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 women were excluded due to intrauterine death, 1 from each group
Selective reporting (reporting bias)	High risk	Maternal and infant outcomes reported. However macrosomia was prespecified but not reported
Other bias	Low risk	No differences in baseline

BMI: body mass index
 GDM: gestational diabetes mellitus
 ITT: intention to treat
 IUGR: intrauterine growth restriction
 LGA: large‐for‐gestational age
 NICU: neonatal intensive care unit
 OGTT: oral glucose tolerance test
 RDS: respiratory distress syndrome
 SGA: small‐for‐gestational age

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Ainuddin 2013	Wrong comparator; metformin versus insulin
Berens 2015	Postpartum intervention
Branch 2010	This study comparing metformin and placebo was registered with ClinicalTrials.gov in 2010. Subsequent updates indicate that the study never started to recruit due to insufficient funding for enrolment of participants
Hebert 2011	This is a pharmacokinetic study and not an intervention study
Smith 2015	Postpartum intervention

Characteristics of studies awaiting assessment [ordered by study ID]

Coiner 2015.

Methods	Unclear if this is a randomised trial
Participants	32 women who had undergone amniocentesis for fetal lung maturity
Interventions	Control ‐ non diabetic Metformin only Glibenclamide only Metformin and insulin
Outcomes	Amniotic levels of insulin, glucose and adiponectin
Notes	We are attempting to identify and contact the authors of this study as it is not clear if the women were randomised to treatment or not prior to the amniocentesis. This appears to be unlikely given that there is a control group of women without diabetes. It is also unclear if these are women with GDM or pregestational diabetes

Sheizaf 2006.

Methods	Parallel, randomised controlled trial
Participants	200 pregnant women Inclusion: women diagnosed with GDM or type 2 diabetes, singleton pregnancy Exclusion: diabetic nephropathy or proliferative retinopathy, unable to swallow tablets Setting: Israel
Interventions	Metformin (no other details) Comparison (not specified)
Outcomes	None prespecified
Notes	This study was first registered in 2006 (NCT00414245) but does not appear to have started recruitment. We contacted investigators in December 2015 to ascertain study status and to find out what the comparison was in the study

GDM: gestational diabetes mellitus

Characteristics of ongoing studies [ordered by study ID]

Moore 2016.

Trial name or title	Glibenclamide (glyburide) versus Glucovance in the treatment of GDM (GGIG)
Methods	Parallel randomised controlled trial
Participants	Inclusion criteria: GDM, > 12 weeks' gestation, able to consent Exclusion criteria: unable to consent, pre‐existing diabetes, glucose‐6‐phosphate dehydrogenase deficiency, serum creatinine > 1, liver disease, allergy to sulfa, allergy to glyburide, allergy to metformin, fetal anomaly
Interventions	Glibenclamide 2.5 mg at bedtime increased to a maximum of 20 mg if needed, versus Glucovance (combination of glibenclamide and metformin) (2.5/500) once daily at bedtime increased if required to 20/2000
Outcomes	Glycaemic control, maternal hypoglycaemia, birthweight, Apgar scores, admission to NICU, neonatal hypoglycaemia
Starting date	June 2016
Contact information	Lisa Moore: lisa.e.moore@ttuhsc.edu
Notes

GDM: gestational diabetes mellitus
 NICU: neonatal intensive care unit

Differences between protocol and review

The following differences are noted between the published protocol and the full review.

Objectives

• Review: To evaluate the effects of oral anti‐diabetic pharmacological therapies for treating women with GDM.

• Protocol: To evaluate the effectiveness of oral agents in treating women with gestational diabetes for improving maternal and fetal health and well‐being.

Types of interventions

In this section we have clarified that the ‘anti‐diabetic agents’ are in fact pharmacological therapies.

Outcomes for use in GRADE/neonatal outcomes

We have edited ‘Neurosensory disability’ to ‘Neurosensory disability in later childhood’.

Contributions of authors

Julie Brown guarantees this review.

Julie Brown wrote the first version of this protocol and review. She contributed to data extraction and data entry.

Ruth Martis has provided comments on drafts of this review.

Brenda Hughes has provided input from a pharmacological perspective on drafts of this review.

Janet Rowan has provided input from a clinical perspective on drafts of this review.

Caroline Crowther has provided clinical and methodological feedback on drafts of this review.

Sources of support

Internal sources

• An internal University department grant, New Zealand.

  An internal University of Auckland department grant from the Liggins Institute has been awarded to Julie Brown to help with the preparation of several Cochrane systematic reviews as part of an overview of systematic reviews for the treatment of women with gestational diabetes. This current protocol/review will be one of the included reviews.

• Liggins Institute, New Zealand.

  Support for infrastructure to support the preparation of this protocol is from the Liggins Institute, University of Auckland, New Zealand.

External sources

• National Institute for Health Research (NIHR), UK.

  NIHR Cochrane Programme Grant Project: 13/89/05 – Pregnancy and childbirth systematic reviews to support clinical guidelines

Declarations of interest

Julie Brown has received a NZD 15,000 internal University grant to assist with the preparation of a Cochrane overview of treatment for gestational diabetes. This is one of the reviews that has been supported. The funds have contributed to a research assistant's time. The funding body gains no financial interest from the publication of the review and has not influenced the content.

Ruth Martis: none known

Brenda Hughes: none known

Janet Rowan: none known

Caroline A Crowther: none known

New