Interventions to prevent women from developing gestational diabetes mellitus: an overview of Cochrane Reviews

Rebecca J Griffith; Jane Alsweiler; Abigail E Moore; Stephen Brown; Philippa Middleton; Emily Shepherd; Caroline A Crowther

doi:10.1002/14651858.CD012394.pub3

. 2020 Jun 11;2020(6):CD012394. doi: 10.1002/14651858.CD012394.pub3

Interventions to prevent women from developing gestational diabetes mellitus: an overview of Cochrane Reviews

Rebecca J Griffith ¹, Jane Alsweiler ¹, Abigail E Moore ², Stephen Brown ³, Philippa Middleton ⁴, Emily Shepherd ⁵, Caroline A Crowther ^2,^✉

Editor: Cochrane Pregnancy and Childbirth Group

PMCID: PMC7388385 PMID: 32526091

Abstract

Background

The prevalence of gestational diabetes mellitus (GDM) is increasing, with approximately 15% of pregnant women affected worldwide, varying by country, ethnicity and diagnostic thresholds. There are associated short‐ and long‐term health risks for women and their babies.

Objectives

We aimed to summarise the evidence from Cochrane systematic reviews on the effects of interventions for preventing GDM.

Methods

We searched the Cochrane Database of Systematic Reviews (6 August 2019) with key words ‘gestational diabetes’ OR ’GDM’ to identify reviews pre‐specifying GDM as an outcome. We included reviews of interventions in women who were pregnant or planning a pregnancy, irrespective of their GDM risk status. Two overview authors independently assessed eligibility, extracted data and assessed quality of evidence using ROBIS and GRADE tools. We assigned interventions to categories with graphic icons to classify the effectiveness of interventions as: clear evidence of benefit or harm (GRADE moderate‐ or high‐quality evidence with a confidence interval (CI) that did not cross the line of no effect); clear evidence of no effect or equivalence (GRADE moderate‐ or high‐quality evidence with a narrow CI crossing the line of no effect); possible benefit or harm (low‐quality evidence with a CI that did not cross the line of no effect or GRADE moderate‐ or high‐quality evidence with a wide CI); or unknown benefit or harm (GRADE low‐quality evidence with a wide CI or very low‐quality evidence).

Main results

We included 11 Cochrane Reviews (71 trials, 23,154 women) with data on GDM. Nine additional reviews pre‐specified GDM as an outcome, but did not identify GDM data in included trials. Ten of the 11 reviews were judged to be at low risk of bias and one review at unclear risk of bias. Interventions assessed included diet, exercise, a combination of diet and exercise, dietary supplements, pharmaceuticals, and management of other health problems in pregnancy. The quality of evidence ranged from high to very low.

Diet

Unknown benefit or harm: there was unknown benefit or harm of dietary advice versus standard care, on the risk of GDM: risk ratio (RR) 0.60, 95% CI 0.35 to 1.04; 5 trials; 1279 women; very low‐quality evidence. There was unknown benefit or harm of a low glycaemic index diet versus a moderate‐high glycaemic index diet on the risk of GDM: RR 0.91, 95% CI 0.63 to 1.31; 4 trials; 912 women; low‐quality evidence.

Exercise

Unknown benefit or harm: there was unknown benefit or harm for exercise interventions versus standard antenatal care on the risk of GDM: RR 1.10, 95% CI 0.66 to 1.84; 3 trials; 826 women; low‐quality evidence.

Diet and exercise combined

Possible benefit: combined diet and exercise interventions during pregnancy versus standard care possibly reduced the risk of GDM: RR 0.85, 95% CI 0.71 to 1.01; 19 trials; 6633 women; moderate‐quality evidence.

Dietary supplements

Clear evidence of no effect: omega‐3 fatty acid supplementation versus none in pregnancy had no effect on the risk of GDM: RR 1.02, 95% CI 0.83 to 1.26; 12 trials; 5235 women; high‐quality evidence.

Possible benefit: myo‐inositol supplementation during pregnancy versus control possibly reduced the risk of GDM: RR 0.43, 95% CI 0.29 to 0.64; 3 trials; 502 women; low‐quality evidence.

Possible benefit: vitamin D supplementation versus placebo or control in pregnancy possibly reduced the risk of GDM: RR 0.51, 95% CI 0.27 to 0.97; 4 trials; 446 women; low‐quality evidence.

Unknown benefit or harm: there was unknown benefit or harm of probiotic with dietary intervention versus placebo with dietary intervention (RR 0.37, 95% CI 0.15 to 0.89; 1 trial; 114 women; very low‐quality evidence), or probiotic with dietary intervention versus control (RR 0.38, 95% CI 0.16 to 0.92; 1 trial; 111 women; very low‐quality evidence) on the risk of GDM. There was unknown benefit or harm of vitamin D + calcium supplementation versus placebo (RR 0.33, 95% CI 0.01 to 7.84; 1 trial; 54 women; very low‐quality evidence) or vitamin D + calcium + other minerals versus calcium + other minerals (RR 0.42, 95% CI 0.10 to 1.73; 1 trial; 1298 women; very low‐quality evidence) on the risk of GDM.

Pharmaceutical

Possible benefit: metformin versus placebo given to obese pregnant women possibly reduced the risk of GDM: RR 0.85, 95% CI 0.61 to 1.19; 3 trials; 892 women; moderate‐quality evidence.

Unknown benefit or harm:eight small trials with low‐ to very low‐quality evidence showed unknown benefit or harm for heparin, aspirin, leukocyte immunisation or IgG given to women with a previous stillbirth on the risk of GDM.

Management of other health issues

Clear evidence of no effect: universal versus risk based screening of pregnant women for thyroid dysfunction had no effect on the risk of GDM: RR 0.93, 95% CI 0.70 to 1.25; 1 trial; 4516 women; moderate‐quality evidence.

Unknown benefit or harm: there was unknown benefit or harm of using fractional exhaled nitrogen oxide versus a clinical algorithm to adjust asthma therapy on the risk of GDM: RR 0.74, 95% CI 0.31 to 1.77; 1 trial; 210 women; low‐quality evidence. There was unknown benefit or harm of pharmacist led multidisciplinary approach to management of maternal asthma versus standard care on the risk of GDM: RR 5.00, 95% CI 0.25 to 99.82; 1 trial; 58 women; low‐quality evidence.

Authors' conclusions

No interventions to prevent GDM in 11 systematic reviews were of clear benefit or harm. A combination of exercise and diet, supplementation with myo‐inositol, supplementation with vitamin D and metformin were of possible benefit in reducing the risk of GDM, but further high‐quality evidence is needed. Omega‐3‐fatty acid supplementation and universal screening for thyroid dysfunction did not alter the risk of GDM. There was insufficient high‐quality evidence to establish the effect on the risk of GDM of diet or exercise alone, probiotics, vitamin D with calcium or other vitamins and minerals, interventions in pregnancy after a previous stillbirth, and different asthma management strategies in pregnancy. There is a lack of trials investigating the effect of interventions prior to or between pregnancies on risk of GDM.

Plain language summary

Interventions to prevent women from developing diabetes during pregnancy: an overview of Cochrane systematic reviews

What is the issue?

Gestational diabetes mellitus (GDM) is defined as high blood glucose levels (hyperglycaemia) first detected during pregnancy. GDM can affect the health of women and their babies.

During pregnancy a woman’s body changes how it processes the nutrients from her food, to ensure that the baby is well nourished. In the first three months the mother has increased sensitivity to insulin. In the second and third trimesters her insulin sensitivity is reduced. Women with GDM have less of an initial increase in sensitivity and their insulin sensitivity is reduced beyond normal later in pregnancy, resulting in the mother developing high blood glucose levels. Her blood levels of fats are also higher than normal, which may contribute to the risk of the baby becoming large for its gestational age.

Why is this important?

Women with GDM are more likely to develop complications in pregnancy including high blood pressure and need labour to be induced. They are at increased risk later of developing type 2 diabetes. Babies born to women with GDM are more likely to be born large, and therefore to experience birth injuries. Once born, the babies are at higher risk of experiencing difficulties in breathing, jaundice and reduced blood sugar levels, and later obesity and diabetes.

There are many risk factors for GDM, making it likely that interventions before/during pregnancy could reduce the risk of women developing GDM. This overview summarises evidence from Cochrane Reviews of randomised controlled trials on interventions that might prevent GDM.

What evidence did we find?

We searched the Cochrane Library (August 2019) and identified 11 Cochrane Reviews that assessed interventions during pregnancy and reported on GDM. The reviews had findings from 71 randomised controlled trials involving 23,154 pregnant women. Interventions included diet, exercise, a combination of diet and exercise, dietary supplements, medications, and management of other health problems. The evidence from the trials ranged from very low to high quality. We identified a further 10 reviews that may provide more information on this topic in the future.

Diet and exercise

Diet and exercise together possibly reduced the risk of a woman developing GDM when compared to standard care (19 trials; 6633 women; moderate‐quality evidence).

Dietary advice alone (5 trials; 1279 women; very low‐quality evidence) and a low glycaemic index diet compared with a moderate to high glycaemic index diet (4 trials; 912 women; low‐quality evidence) had an unclear effect on the risk of GDM. Exercise alone had an unclear effect on the risk of GDM (3 trials; 826 women; low‐quality evidence).

Dietary supplements

Omega‐3 fatty acid supplementation in pregnancy had no effect (12 trials; 5235 women; high‐quality evidence).

Myo‐inositol supplementation during pregnancy possibly reduced the risk of GDM (3 trials with 502 women; low‐quality evidence).

Vitamin D supplementation in pregnancy had a possible benefit in reducing the risk of developing GDM (4 trials with 446 women; low‐quality evidence). These trials were all from Asian countries and the women’s vitamin D levels before supplementation were mostly unknown.

Vitamin D given with calcium supplementation, or with calcium plus other minerals had an unclear effect.

Probiotic with dietary intervention had an unclear effect on the risk of developing GDM.

Medications

The drug metformin had a possible benefit in reducing the risk of developing GDM when given to obese pregnant women (3 trials; 892 women; moderate‐quality evidence).

Low‐ to very low‐quality evidence from eight small trials showed unclear effect on GDM risk for heparin, aspirin, leukocyte immunisation or immunoglobulin (IgG) given to women who had previously experienced a stillbirth.

Management of other health issues

Universal versus risk‐based screening for thyroid problems had no effect on the risk of GDM (1 trial; 4516 women; moderate‐quality evidence). Two different approaches to management of the mothers’ asthma had an unclear effect (low‐quality evidence).

What does this mean?

A combination of exercise and diet, supplementation with myo‐inositol and vitamin D supplementation were of possible benefit in reducing the risk of developing GDM. Further high‐quality evidence from randomised controlled trials is needed to confirm these results, and to look further at the use of metformin. No trials assessed interventions before pregnancy.

Background

Description of the condition

Gestational diabetes mellitus (GDM) is glucose intolerance causing hyperglycaemia with onset during pregnancy. The optimal blood glucose concentration cut‐off to diagnose GDM remains controversial (ACOG 2018; Cheung 2018; Coustan 2010; HAPO 2008; IADPSG 2010; Ministry of Health 2014; NICE 2015). Lower blood glucose concentration thresholds for diagnosis have resulted in more women being diagnosed with GDM (IADPSG 2010; Nankervis 2014; WHO 2013). The prevalence of GDM varies internationally with approximately 15% of pregnant women affected (Bottalico 2007; Egan 2017; Ferrara 2007; Guariguata 2014; McIntyre 2018; Melchior 2017; NICE 2015). Prevalence varies by country, ethnicity and the diagnostic thresholds (Farrar 2016; HAPO 2008; Pu 2015).

Normal pregnancy is characterised by changes in the metabolism of carbohydrate, amino acids and lipids. The result of these combined changes is a switch to maternal use of lipids as a source of energy, sparing glucose and amino acids for the fetus.

In the first trimester of pregnancy there is increased insulin sensitivity (Catalano 1991; Catalano 1992; Catalano 1993) as a result of adaptation of the pancreatic ß‐cells (Van Assche 1978), increased insulin synthesis (Weinhaus 1996) and secretion (Sorenson 1993), leading to improved utilisation and oxidation of glucose. By 14 weeks’ gestation, the first phase of insulin secretion in response to a glucose load has increased by approximately 120% (Bowes 1996; Catalano 1993), resulting in a reduced plasma glucose concentration (Catalano 1992). In the second and third trimesters, insulin sensitivity reduces (Catalano 1991; Catalano 1992; Catalano 1993; Ryan 1985), and hepatic gluconeogenesis increases (Nelson 1994).

In women with GDM, the increase in first phase insulin secretion in response to a glucose load is reduced (Catalano 1993), with inadequate adaptation of ß‐cells leading to impaired glucose homeostasis (Buchanan 2001). In the second and third trimesters the insulin sensitivity is further reduced (Catalano 1998) with resultant maternal hyperglycaemia, elevated glycated haemoglobin (HbA1c) concentrations and increased transport of glucose across the placenta to the developing fetus (Setji 2005).

During normal pregnancy, changes in adipocytes result in more fat stores being laid down with increased synthesis and reduction in clearance of triglycerides (Ginci 1997), resulting in an increase in all plasma lipid components (Butte 2000). Women with GDM have hyperlipidaemia (raised levels of lipids in the blood) beyond that seen in normal pregnancy (Koukkou 1996) due to increased synthesis in the liver and reduced activity of lipoprotein lipase and hepatic lipase (Ginci 1997; Sattar 1997). Higher concentrations of free fatty acids cross the placenta than in normal pregnancy (Larqué 2011), which may contribute to the risk of macrosomia (large baby) (Knopp 1992).

There are multiple risk factors for GDM (Berkowitz 1992;Chu 2007; Khan 2013;Pu 2015; Solomon 1997;Theriault 2014;Xiong 2001). These include advanced maternal age, maternal high or low birthweight, high parity (Petry 2010), polycystic ovarian syndrome (Toulis 2009), a past history of GDM (Kim 2007), family history of first‐degree relatives with GDM or type 2 diabetes (Petry 2010), maternal overweight or obesity (Morisset 2010; Torloni 2009), physical inactivity before or in early pregnancy (Dempsey 2004; Tobias 2011; Zhang 2014), gestational weight gain (Morisset 2010), and a past history of a macrosomic baby or a stillbirth (Petry 2010). Impaired glucose tolerance is common to many of these risk factors, but the exact mechanisms by which each contributes to the development of GDM are uncertain. Some of these risk factors are potentially modifiable through preventive interventions.

Women with GDM have a higher risk of pre‐eclampsia and need for induction of labour (Dodd 2007). Women with a history of GDM have a greater than seven‐fold increased risk of developing type 2 diabetes later, with more than half these women developing type 2 diabetes within 10 years after giving birth (Bellamy 2009; Kim 2002). Infants born to mothers with GDM are at increased risk of being born large‐for‐gestational age (Sacks 2015), and are therefore more likely to experience birth injuries such as nerve palsy, bone fracture and shoulder dystocia. In the neonatal period, they are at higher risk of respiratory distress syndrome, jaundice and hypoglycaemia (reduced levels of blood sugar) (Adams 1998; Crowther 2005; Gonzalez‐Quintero 2007; He 2015; Landon 2009; Langer 2005). Longer‐term health consequences into childhood and adulthood include obesity, diabetes, the metabolic syndrome (Boney 2005; Cho 2000), and adverse neurodevelopmental outcomes (Chatzi 2014; Fraser 2012; Nelson 2000; Torres‐Espinola 2015).

Description of the interventions

Interventions to prevent GDM have been used preconception, during pregnancy and inter‐conception (between pregnancies). Preconception and inter‐conception interventions have been used, particularly in women at high risk of GDM, such as those who are overweight or obese (Yeung 2010), or with a history of GDM (Khan 2013; Shyam 2013). The opportunity exists to intervene with health promotion strategies prior to and between pregnancies for women identified with risk factors for GDM (Hanson 2015; Jack 1990).

Interventions directed at preventing GDM include dietary or exercise interventions, or a combination of these, dietary supplement interventions and pharmaceutical interventions.

The focus of some dietary advice interventions for GDM prevention have been specific, such as increasing fibre intake (Fraser 1983; Fraser 1988) or aiming for a low glycaemic index diet (Kizirian 2017; Markovic 2015). Others have included broader advice regarding “healthy eating” as part of more comprehensive lifestyle interventions (Quinlivan 2011).

Exercise or physical activity interventions for preventing GDM have varied from general advice to specific individualised programs using a range of different activities, such as aerobic activities, stationary cycling or yoga (Barakat 2012; Guelfi 2016; Ong 2009; Rakhshani 2012; Stafne 2012). These have been employed in isolation (Barakat 2012; da Silva 2017; Goodarzi‐Khoigani 2017; Guelfi 2016; Ong 2009; Stafne 2012), or in combination with dietary interventions (Dodd 2014; Harrison 2013; Koivusalo 2016; Luoto 2011; Petrella 2014; Poston 2015; Simmons 2015).

Dietary supplement interventions such as probiotics (Lindsay 2014; Luoto 2010; Wickens 2017), myo‐inositol (D'Anna 2013; Farren 2017; Santamaria 2016), vitamin D (Bao 2017; Soheilykhah 2013), and fish oils have been investigated for GDM prevention (Zhou 2012).

Pharmaceutical therapies, which may have a role in GDM prevention, include sulphonylureas, biguanides, thiazolidinediones, alpha‐glucosidase inhibitors, meglitinides and peptide analogues. Metformin or glibenclamide (also known as glyburide) are the only oral hypoglycaemics recommended in clinical practice guidelines for use in pregnancy (ACOG 2018; NICE 2015). However, there is a paucity of data regarding the safety of many of these in pregnancy (Holt 2014; Kavitha 2013; Slocum 2002). Where safety data are available, this has often been limited to short‐term health outcomes. The Metformin in Gestational diabetes (MiG) trial investigated the use of metformin for women with GDM, and demonstrated metformin was safe in the offspring, at least up to nine years of age (Rowan 2018).

How the intervention might work

Dietary interventions

Different dietary components have direct and indirect effects on glycaemic profile. Interventions that alter these could be utilised to reduce GDM risk (Ley 2011; Rogozinska 2015; Zhang 2006). Insulin sensitivity and secretion are reduced in association with high simple sugar intake (Davis 2005; Reiser 1979). With ongoing high sugar intake, pancreatic exhaustion may ensue with impaired glucose tolerance (Ludwig 2002). Less insulin resistance is seen with a high‐fibre and low‐glycaemic index diet. Fibre slows digestion (Burton‐Freeman 2000; Jenkins 2000; Vahouny 1988) and rate of glucose absorption, thus altering the blood glucose concentration and insulin response (Jenkins 2000; Liese 2005; Mcintosh 2001). Increasing dietary fibre intake may reduce appetite and hence insulin resistance associated with adiposity (Burton‐Freeman 2000). Intake of protein and fats may also reduce appetite (Tannous dit El Khoury 2006) with similar effect (Kantartzis 2009; Kohrt 1993; Pan 1993), and through protection of ß‐cells from oxidative injury (Cai 2012; Lin 2012). A general reduction in calorie intake with resultant weight loss and reduced adiposity improves insulin sensitivity and glycaemic profile (Knopp 1991; Larson‐Meyer 2006). This needs to balanced against potential risks of weight loss during pregnancy such as ketonaemia associated with marked calorie restriction (Churchill 1969;Magee 1990;Metzger 2007; Ornoy 1998;Rizzo 1991). Due to these concerns, international guidelines do not recommend hypocaloric diets during pregnancy (Ireland 2010; NICE 2010; Thompson 2013). There is ongoing debate as to whether calorie restriction might be appropriate in overweight and obese pregnant women (Knopp 1991; Procter 2014). Interventions to aid with weight loss in overweight or obese women preconception or inter‐conception may reduce the risk of GDM in any future pregnancy (Pole 1999).

Exercise interventions

The risk of developing GDM is inversely associated with the amount of regular physical activity before or during pregnancy (Dempsey 2004; Zhang 2014). There is increased energy expenditure and hence glucose consumption during exercise; blood flow through muscle mass and capillary surface area for glucose exchange increases (Rose 2005; Sjøberg 2011). During muscle contraction, there is translocation of the glucose transporter type 4 (GLUT‐4) from within skeletal muscle cells to the surface (Jessen 2005; Kennedy 1999; Rose 2005) with resultant increased glucose uptake. The increase in insulin sensitivity continues beyond the exercise period (Jensen 2012; Perseghin 1996). Muscle mass increases with regular physical activity, and thus glucose tolerance and insulin sensitivity are likely to improve (Yki‐Jarvinen 1983).

Diet and exercise interventions combined

Combined interventions targeting more than one of the multiple risk factors for GDM could be synergistic. Prevention of type 2 diabetes has been demonstrated using combined dietary, exercise and weight loss interventions (Haw 2017; Knowler 2002; Tuomilehto 2001). These might be expected to have a similar effect in prevention of GDM.

Dietary supplement interventions

Probiotics

Probiotic use can change the microbiome of the gut, which may reduce insulin resistance (FAO/WHO 2001; Hill 2014; Kondo 2010), through decreasing inflammatory signalling (Ma 2008), and upregulating genes involved in fat metabolism and insulin sensitivity (Kondo 2010).

Myo‐inositol

Myo‐inositol, a polyol with insulin‐mimetic properties (Croze 2013; Saltiel 1990) involved in insulin transduction signalling (Baillargeon 2010), increases GLUT‐4 translocation to the cell membrane in skeletal muscle (Dang 2010), thus improving insulin sensitivity (Corrado 2011). Supplementary use in polycystic ovarian syndrome results in reduced fasting insulin concentrations (Unfer 2017). Myo‐inositol is present in the diet in some seeds, grains, nuts, beans, vegetables and fruit (Clements 1980).

Vitamin D

Vitamin D deficiency is associated with insulin resistance (Esteghamati 2014) and poor pancreatic β‐cell function (Chiu 2004). Vitamin D may affect insulin secretion by binding to vitamin D receptors in the pancreas and regulating the balance between the extracellular and intracellular calcium pools (Sooy 1999). Vitamin D deficiency may reduce pancreatic insulin secretion (Bourlon 1999; Norman 1980), while supplementation with vitamin D may influence the expression of insulin‐sensitive genes (Alkharfy 2013), thus reducing inflammatory markers and improving glucose uptake (Marcotorchino 2012).

Fish oil

The circulating concentrations of several long‐chain polyunsaturated fatty acids are altered in GDM (Wijendran 1999). Omega‐3 fatty acids have several anti‐inflammatory effects (Calder 2006). The predominant sources of the omega‐3 fatty acids, eicosapentaenoic acid and docosahexaenoic acid, are fish and fish oils (Kris‐Etherton 2000; Kris‐Etherton 2009). The lipid composition of cell membranes is altered with changes to dietary fatty acid composition (Calder 2006; Lardinois 1987), which affects insulin binding and sensitivity. Increased insulin secretion and sensitivity may result from omega‐3 or fish oil supplementation (Baynes 2018). Increased inflammation can result in insulin resistance, while omega‐3 fatty acids inhibit TLR‐2 and TLR‐4 receptors for inflammatory cytokines (Lee 2004).

Pharmaceutical interventions

Metformin, a biguanide, crosses the placenta with similar concentrations found in maternal and fetal circulations (Vanky 2005). Metformin reduces hepatic gluconeogenesis (Stumvoll 1995; Wollen 1988), and enhances peripheral glucose uptake and utilisation (Viollet 2012), improving insulin sensitivity and reducing hyperglycaemia (Jackson 1987). Metformin may enhance insulin sensitivity and preserve pancreatic β‐cell capacity in women with polycystic ovarian syndrome (Ainuddin 2015). Glibenclamide crosses the placenta with fetal blood concentrations approximately 70% of maternal blood concentrations (Hebert 2009). Glibenclamide stimulates insulin secretion, but has been associated with an increased risk of macrosomia, neonatal hypoglycaemia and higher maternal weight gain in comparison to metformin (Liang 2017).

Why it is important to do this overview

A number of risk factors for GDM, such as physical inactivity, being overweight or obese prior to pregnancy, and having a poor diet are potentially modifiable. While different strategies have shown promise in the prevention of GDM, it is currently unclear which strategies are most effective. Primary prevention of GDM rather than treatment would lead to economic (Danyliv 2014) and health benefits.

This overview provides an important resource for all healthcare professionals caring for pregnant women, guideline developers, policy makers, researchers, and pregnant women at risk of developing GDM, and their families.

Use of the overview to identify and target effective preventive interventions may contribute to reducing the increase in rates of GDM seen globally, thus reducing the significant short‐ and long‐term health risks for the mothers and their infants. Further, this overview identifies priority areas requiring further research.

Objectives

We aimed to summarise the evidence from Cochrane systematic reviews on the effects of interventions for preventing gestational diabetes mellitus (GDM).

Methods

Criteria for considering reviews for inclusion

In this overview of systematic reviews, we included only Cochrane systematic reviews, that assessed interventions that may prevent gestational diabetes mellitus (GDM), reporting GDM as a primary or secondary outcome of the review. We identified Cochrane protocols and titles for potential future inclusion in an update of the overview, and classified them as ’ongoing reviews’ (in Appendix 1).

When reviews were identified for inclusion that were more than two years out of date, we contacted the Cochrane Pregnancy and Childbirth Editorial Base to identify whether an update was in progress. We did not contact individual review authors, but noted the publications and search dates of the reviews. When a review was out of date and would not be updated in time to be included in the overview, we included the last published version and acknowledged this as a potential limitation.

Participants

We included women planning a pregnancy, between pregnancies, or pregnant women. We excluded women with pre‐existing type 1 or type 2 diabetes.

Interventions

We included interventions prior to pregnancy (preconception), between pregnancies (inter‐conception), or implemented prior to GDM screening in pregnancy, and for each of these time periods these included:

dietary interventions;
exercise interventions;
dietary and exercise interventions combined;
dietary supplement interventions (e.g. probiotics, myo‐inositol, vitamin D and omega‐3 fatty acids);
pharmaceutical interventions (e.g. oral anti‐diabetic pharmaceutical therapies);
interventions for the management of other health problems during pregnancy.

We included reviews comparing the above interventions with standard care or no intervention (or a placebo), as well as those comparing different interventions.

Outcome

GDM (defined by review authors and trialists).

Search methods for identification of reviews

We searched the Cochrane Database of Systematic Reviews (6 August 2019) using key words ‘gestational diabetes’ OR ’GDM’. We used the search terms to search ’all text’, and did not limit to ’title, abstract, or keywords’. We did not apply any language or date restrictions. The review group was contacted to identify titles for future inclusion and no additional titles were identified.

Data collection and analysis

We based the methodology for data collection and synthesis on Chapter 22, 'Overviews of Reviews' in the Cochrane Handbook of Systematic Reviews of Interventions (Higgins 2011).

Selection of reviews

Two overview authors independently assessed for potential inclusion all Cochrane systematic reviews we identified that evaluated the effects of the aforementioned interventions and reported on the incidence of GDM. We then assessed the methods section of reviews and protocols to ensure those with the appropriate population and pre‐specified outcome were selected for inclusion. We resolved any disagreement through discussion or consultation with a third overview author.

Data extraction and management

Two of the overview authors independently extracted data from each systematic review, using an electronic form which we designed and piloted. We resolved disagreements by consensus or by discussion with a third overview author. When information from the review was missing, we accessed the published papers of the individual study and contacted the systematic review authors for further details. We extracted and tabulated information for the following.

Review characteristics

Review title and authors.
Search date: date of search conducted by review (we considered less than two years ago to be current).
The number of trials in the review, number of women and their infants, and their characteristics.
Risk of bias of the included trials (as reported by the review authors; see ’Risk of bias of included studies within reviews’ below, under Assessment of methodological quality of included reviews).
Interventions and comparisons relevant to this overview.
The prespecified outcome (GDM) relevant to this overview.
Any other characteristics required to assess and report on review quality (see 'Quality of included reviews' under Assessment of methodological quality of included reviews).

Statistical summaries

The summary intervention effects, including the pooled effects (e.g. risk ratios (RRs), odds ratios (ORs), mean differences (MDs), 95% confidence intervals (CIs)), and numbers of studies and participants contributing data to each pooled effect from comparisons and for the outcome relevant to this overview.
Results of any subgroup or sensitivity analyses conducted by the authors for overview outcome.
Information required to assess and report on the quality of the evidence for the intervention effects extracted above (see ’Quality of evidence in included reviews’ under Assessment of methodological quality of included reviews).

All reviews performed meta‐analyses for GDM, the outcome for our overview.

Assessment of methodological quality of included reviews

Quality of included reviews

Two overview authors assessed for relevance, by checking that the population, intervention, comparator and outcomes aligned between the review and overview, and independently assessed the methodological quality of the included systematic reviews using the Risk of Bias in Systematic Reviews (ROBIS) tool (Whiting 2015). We resolved differences through discussion and, when needed, through discussion with a third overview author.

The ROBIS tool assesses risk of bias across four domains.

Study eligibility criteria.
Identification and selection of studies.
Data collection and study appraisal.
Synthesis and findings.

ROBIS uses signalling questions to assess specific concerns about potential biases within the review, and the ratings from these questions are used to judge overall risk of bias. The signalling questions were answered as ’yes’, ’probably yes’, ’probably no’, ’no’ or ’no information’. Each of these four domains were then designated as low, high or unclear risk of bias. If the answers to all signalling questions for a domain were ’yes’ or ’probably yes’, the level of concern could be judged as low. If any signalling question was answered ’no’ or ’probably no’, the potential for concern about bias exists. Finally, a summary judgement based on any concerns, assessment of methods and bias was made of low, high or unclear overall risk of bias for the systematic review.

Quality of evidence in the included reviews

We assessed the quality of the evidence for GDM using GRADE (Grades of Recommendation, Assessment, Development and Evaluation). When available, we used the GRADE assessments from the included Cochrane systematic reviews. When this was not available, we used the GRADE system to review pooled summary statistics and risk of bias of included trials. The GRADE system assesses the following features for the evidence found for selected outcomes.

Risk of bias: internal validity of the evidence.
Inconsistency: heterogeneity or variability in the estimates of effect across studies.
Indirectness: degree of differences between population, intervention and outcome of interest.
Imprecision (random error): extent to which confidence in the effect estimate is adequate to support a particular decision.
Risk of publication bias: degree of selective publication of studies.

The GRADE system rates the quality of the evidence as:

high (further research is very unlikely to change confidence in the estimate of the effect);
moderate (further research is likely to have an important impact on confidence in the estimate of effect and may change the estimate);
low (further research is very likely to have an important impact on confidence in the estimate of effect and is likely to change the estimate);
very low (any estimate of effect is very uncertain).

We summarised the evidence in 'Summary of findings' tables which we populated with the summary risk estimates and 95% CIs, number of participants, and the quality of the review for each intervention and whether GDM was a primary or secondary review outcome. We planned to include timing of intervention (preconception, inter‐conception and during pregnancy), but all interventions within our included reviews were during pregnancy.

Risk of bias of included studies within reviews

We did not reassess the risk of bias of included studies within reviews, but instead reported study risk of bias according to the review authors’ assessment. In the case that individual studies were included in two or more Cochrane Reviews, we report this, and any variation regarding review authors’ assessments of study risk of bias. We collected this information during the data extraction process.

Data synthesis

We undertook a narrative description of the included Cochrane systematic reviews. We did not examine indirect comparisons or conduct network meta‐analyses. We summarised the main results of the included reviews by categorising their findings in the following framework, and by intervention focus/topic. We planned to organise by timing of intervention (preconception, inter‐conception and during pregnancy), but no preconception or inter‐conception interventions were identified. We assigned graphic icons to communicate the direction of review effect estimates and our confidence in the available data. This is the framework adopted by Medley and colleagues in their overview on 'Interventions during pregnancy to prevent preterm birth: an overview of Cochrane systematic reviews' (Medley 2018), and was based on graphics produced by the World Health Organization to describe different types of workers and their roles in maternal and newborn care (http://optimizemnh.org/optimizing-health-worker-roles-maternal-newborn-health/). We used graphic icons to indicate mutually exclusive assessment categories (see Figure 1).

Clear evidence of benefit (moderate‐ or high‐quality evidence with CIs not crossing line of no effect).
Clear evidence of harm (moderate‐ or high‐quality evidence with CIs not crossing line of no effect).
Clear evidence of no effect or equivalence (moderate‐ or high‐quality evidence with narrow CIs crossing the line of no effect).
Possible benefit (low‐quality evidence with clear benefit, or moderate‐ or high‐quality evidence with wide CIs crossing the line of no effect).
Possible harm (low‐quality evidence with clear harm, or moderate‐ or high‐quality evidence with wide CIs crossing the line of no effect).
Unknown benefit or harm (low‐quality evidence with wide CIs crossing the line of no effect or very low‐quality evidence).

The choice of category reflected the information synthesised from the included reviews for the overview outcome (GDM). We used separate assessments for different comparisons when required (e.g. where one intervention was compared with both placebo (or no treatment) and with an alternative intervention). This approach to summarising the evidence is based on an earlier overview ‘Antenatal and intrapartum interventions for preventing cerebral palsy: an overview of Cochrane systematic reviews’ (Shepherd 2017).

We conducted a sensitivity analysis excluding Cochrane systematic reviews with a ROBIS (Risk of Bias in Systematic Reviews) review rating that was of high concern for risk of bias in any domain.

Results

Our search of the Cochrane Database of Systematic Reviews identified 129 results. After excluding six overviews, two protocols which had been withdrawn, and one review which had been superceded, we searched the text of 120 protocols and completed systematic reviews for the outcome, GDM. We assessed the full text of those protocols and reviews which pre‐specified GDM as an outcome. No overview author assessed their own systematic review (several eligible reviews were authored by members of the overview team).

We included 11 reviews in this overview. We excluded 98 reviews that included participants with pre‐existing diabetes or did not pre‐specify GDM as a review outcome and one review (Crowe 2019) for which there was no comparison group (see Figure 2).

We listed 10 protocols and reviews, that may provide data in future updates of this overview, in the appendices.

Appendix 1, Ongoing reviews, cites one Cochrane protocol that pre‐specifies GDM as a secondary outcome, and will be considered for inclusion in future updates of this overview when it is published as a full review.
Appendix 2, Reviews awaiting further classification, summarises nine Cochrane Reviews that pre‐specified GDM as a primary or secondary outcome, but the included trials had no data reported for this outcome. These reviews will be considered again for inclusion in future updates of this overview.

Description of included reviews

The 11 included systematic reviews all involved pregnant women. They included 71 trials of 23,154 women. There was one review for each of the following:

dietary advice versus standard care or different dietary advice (Tieu 2017);
any type of exercise regimen versus standard care or another type of exercise regimen (Han 2012);
any combined dietary and exercise combination versus standard care or a different dietary and exercise combination (Shepherd 2017);
omega‐3 fatty acid supplementation (supplement or food) versus placebo or no omega‐3 (Middleton 2018);
myo‐inositol versus control or placebo (Crawford 2015);
vitamin D (alone or in combination with other micronutrients) versus other micronutrients or control (Palacios 2019);
probiotics versus "any type" of comparison intervention (Barrett 2014);
metformin (alone or in combination) versus placebo or control in overweight or obese women (Dodd 2018);
any single intervention, combination of interventions or tailored model of care/algorithm/guideline/protocol for improving health outcomes in subsequent pregnancies following stillbirth versus no intervention or standard care (Wojcieszek 2018);
screening for thyroid dysfunction versus no screening or an alternative screening method for improving maternal and infant health (Spencer 2015);
interventions for managing asthma in pregnancy versus alternative interventions, or placebo or control (Bain 2014).

All the included reviews reported interventions during pregnancy. No reviews reported interventions prior to, or between pregnancies. The number of randomised trials that reported data on GDM in the included systematic reviews ranged from one to 19. The number of pregnant women in the included reviews ranged from 217 (Wojcieszek 2018) to 6633 (Spencer 2015). Only one review included trials from low‐income countries (Palacios 2019). Six reviews were published more than two years ago (Bain 2014; Barrett 2014; Crawford 2015; Han 2012; Spencer 2015; Tieu 2017). One review (Shepherd 2017), though published within the last two years had conducted the search more than two years ago. Four reviews (Dodd 2018; Middleton 2018:Palacios 2019; Wojcieszek 2018) had conducted searches in the last two years and were considered up to date. Han 2012 is undergoing an update. There are no known active updates in progress for the remaining reviews.

Further details of the included reviews can be found in Table 1 'Characteristics of included reviews'. Details of the 'Risk of bias' assessments included in each review can be found in Table 2.

1. Characteristics of included reviews.

Review ID	Date of search; date assessed as up‐to‐date	No. included trials; years of publication and countries	No. participants in included trials	No. participants and trials with data on GDM	Inclusion criteria for “Types of participants”	Relevant comparison interventions (no. trials and participants)	GDM primary or secondary outcome in review
Bain 2014	Search June 2014; Up‐to‐date October 2014	8 RCTs published in 1996, 1998, 2004, 2005, 2006, 2011, 2012 (2) in Brazil, USA (2), Egypt, Australia, Germany and multi country	1181	268 (2)	Pregnant women with current asthma (with a health professional's diagnosis), regardless of age, parity, plurality, and severity of asthma	Any intervention to manage asthma (pharmacological/non‐pharmacological) versus placebo/no intervention/alternative intervention	Secondary
Barrett 2014	Search August 2013; Up‐to‐date February 2014	1 RCT in 2008 in Finland	256	256 (1)	Pregnant women including those with previous GDM	Probiotic in combination with diet vs diet with placebo or placebo alone	Primary
Crawford 2015	Search November 2015; Up‐to‐date December 2015	4 RCTs in 2013 (2), 2014, 2015 in Italy	567	502 (3)	Pregnant women	Any doses of myo‐inositol in pregnancy, alone or in a combination preparation versus no treatment or placebo or another intervention	Primary
Dodd 2018	Search October 2017; Up‐to‐date July 2018	3 RCTs in 2015 (2), 2016 in Egypt and UK (2)	1099 (data for 1034)	892 (3)	Pregnant women with obesity or who are overweight, defined as women with booking or early pregnancy or pre‐pregnancy body mass index (BMI) ≥ 25.0 kg/m2	Metformin versus placebo or no metformin	Secondary
Han 2012	Search April 2012; Up‐to‐date July 2012	5 RCTs in 2009, 2010 (2), 2011, 2012 in Australia, (2), Norway, Spain, New Zealand	1115 (data for 922)	826 (3)	Pregnant women of any age, gestation, parity or plurality	Any types of exercise and lifestyle management versus standard care or different frequencies of same intervention or different interventions	Primary
Middleton 2018	Search 16 August 2018; up to date 14 November 2018	70 RCTs. 1989‐2018. Multiple countries including Bangladesh, Mexico, Venezuela	19,927	5235 (12)	Pregnant women, regardless of their risk for pre‐eclampsia, preterm birth or intrauterine growth restriction.	Omega‐3 fatty acids (usually fish or algal oils) compared with placebo or no omega‐3 fatty acids. Trials that assessed omega‐3 fatty acid co‐interventions (e.g. omega‐3 with another agent) Studies or study arms that compared omega‐3 doses or types of omega‐3 (e.g. DHA versus EPA) directly	Secondary
Palacios 2019	Search July 2018; Up‐to‐date July 2019	30 RCTs in 1980 (1), 1986 (2), 1987 (1), 1988 (1), 1991 (1), 2000 (1), 2002 (1), 2008 (1), 2009 (1), 2010 (1), 2011 (1), 2012 (3), 2013 (6), 2014 (1), 2015 (2), 2016 (4), 2017 (2) in Australia (1), Bangladesh (2), Brazil (1), China (1), France (2), India (5), Iran (12), New Zealand (1), Pakistan (1), Russia (1) and the UK (3)	7033	446 (1)	Pregnant women any gestation or chronological age, parity, number of fetuses	Vitamin D (any dose, duration, time of commencement) during pregnancy (alone or in combination with other micronutrients) versus placebo/no intervention/other vitamins or minerals	Primary
Shepherd 2017	Search November 2016; Up‐to‐date November 2017	23 RCTs (2 cluster and 21 individually randomised) in 2002, 2009, 2011 (4), 2012, 2013 (5), 2014 (3), 2015 (3), 2016 (4), 2017 in Australia (2), Brazil, Canada (2), China (2), Denmark, Egypt, Finland (3), Germany, Italy (2), Norway, UK (2), USA (5)	8918	6633 (19)	Pregnant women regardless of age, gestation, parity or plurality	Any type of diet intervention with any type of exercise intervention versus no intervention or a different diet and exercise intervention	Primary
Spencer 2015	Search July 2015; Up‐to‐date September 2015	2 RCTs in 2010, 2012 in Italy, UK	26,408	4516 (1)	Women, either pre‐pregnancy or during pregnancy, including both singleton and multiple pregnancies	Any screening method (e.g. tool, program, guideline or protocol) for detecting thyroid dysfunction pre‐pregnancy or during pregnancy versus no screening	Secondary
Tieu 2017	Search January 2016; Up‐to‐date January 2017	11 RCTs (1 quasi randomised) in 1983, 1998, 2006, 2008, 2009 (2), 2011 (2), 2012, 2014, 2016 in Australia (4), Brazil, Denmark, Ireland, Finland, UK, USA (2)	2786	2191 (9)	Pregnant women of any age gestation, parity or plurality	Dietary advice before testing for GDM versus no advice or different types of advice	Primary
Wojcieszek 2018	Search June 2018; Up‐to‐date December 2018	9 RCTs and 1 quasi randomised in 1964, 1994, 1995, 2002, 2004, 2009, 2012, 2014, 2015, 2016 in Canada (2), Denmark (3), France, Israel, Italy, and Pakistan. One trial was undertaken across both Austria and Germany	222	210 (8)	"Parents" who had experienced a stillbirth of 20 weeks’ gestation or more who were pregnant or considering a subsequent pregnancy.	Any single intervention, combination of interventions or tailored model of care/algorithm/guideline/protocol for improving health outcomes in subsequent pregnancies following stillbirth	Secondary

Open in a new tab

Abbreviations:

GDM: gestational diabetes mellitus

RCT: randomised controlled trial

DHA: docosahexaenoic acid

EPA: eicosapentaenoic acid

2. Risk of bias of trials in included reviews.

Review ID	Summary of trial limitations (risk of bias)
Bain 2014	Sequence generation: 5 low, 3 unclear Allocation concealment: 4 low, 4 unclear Blinding (participants and personnel): 3 low, 3 unclear, 2 high Blinding (outcome assessors): 5 low, 3 unclear Incomplete outcome data: 4 low, 4 unclear Selective reporting: 2 low, 5 unclear, 1 high Other: 3 low, 5 unclear Overall: "Overall we judged two trials to be at a low risk of bias, two trials to be at an unclear risk of bias, and the other four trials to be at a moderate risk of bias."
Barrett 2014	Sequence generation: 1 low Allocation concealment: 1 low Blinding (participants and personnel): 1 low Blinding (outcome assessors): 1 low Incomplete outcome data: 1 low Selective reporting: 1 low Other: 1 low Overall: "The included study was assessed to be at low risk of bias across all domains."
Crawford 2015	Sequence generation: 3 low, 1 unclear Allocation concealment: 2 low, 2 unclear Blinding (participants and personnel): 2 unclear, 2 high Blinding (outcome assessors): 1 low, 3 unclear Incomplete outcome data: 2 low, 1 unclear, 1 high Selective reporting: 2 low, 1 unclear, 1 high Other: 2 low, 1 unclear, 1 high Overall: "Overall, there was unclear risk of bias due to insufficient information provided to enable a judgement of risk, particularly with regard to allocation concealment and blinding of outcome assessment."
Palacios 2019	Sequence generation: 21 low 8 unclear 1 high Allocation concealment: 13 low 16 unclear 1 high Blinding (participants and personnel): 15 low 1 unclear 14 high Blinding (outcome assessors): 10 low 18 unclear 2 high Incomplete outcome data: 16 low 4 unclear 10 high Selective reporting: 29 unclear 1 high Other: 14 low 2 unclear 4 high Overall: "The risk of bias was high for allocation and/or blinding in 14 trials and for attrition in 10 trials."
Dodd 2018	Sequence generation: 2 low, 1 unclear Allocation concealment: 2 low, 1 unclear Blinding (participants and personnel): 2 low, 1 unclear Blinding (outcome assessors): 2 low, 1 unclear Incomplete outcome data: 3 low Selective reporting: 2 low, 1 high Other: 2 low, 1 unclear Overall risk of bias reported: no comment in main text
Han 2012	Sequence generation: 2 low, 3 unclear Allocation concealment: 2 low, 3 unclear Blinding (participants and personnel): 5 high Blinding (outcome assessors): 5 unclear Incomplete outcome data: 2 low, 3 high Selective reporting: 3 low, 2 high Other: 3 low, 2 high Overall: "Overall, the risk of bias was judged to be moderate."
Middleton 2018	Sequence generation: 37 low, 32 unclear, 1 high Allocation concealment: 29 low, 40 unclear, 1 high Blinding (participants and personnel): 52 low, 15 unclear, 3 high Blinding (outcome assessors): 39 low, 41 unclear Incomplete outcome data: 13 low, 30 unclear, 27 high Selective reporting: 13 low, 45 unclear, 12 high Other: 34 low, 34 unclear, 2 high Overall: “Overall study‐level risk of bias was mixed, with selection and performance bias mostly at low risk, but there was high risk of attrition bias in some trials.”
Shepherd 2017	Sequence generation: 17 low, 6 unclear Allocation concealment: 13 low, 10 unclear Blinding (participants and personnel): 23 high Blinding (outcome assessors): 8 low, 15 unclear Incomplete outcome data: 12 low, 7 unclear, 4 high Selective reporting: 3 low, 15 unclear, 5 high Other: 16 low, 6 unclear, 1 high Overall: "Primarily due to lack of reporting, the overall risk of bias was judged to be unclear."
Spencer 2015	Sequence generation: 2 low Allocation concealment: 2 low Blinding (participants and personnel): 1 low, 1 unclear Blinding (outcome assessors): 2 low Incomplete outcome data: 1 low, 1 unclear Selective reporting: 2 unclear Other: 2 low Overall: "Overall, the two trials were judged to be of low risk of bias."
Tieu 2017	Sequence generation: 7 low, 2 unclear, 1 high Allocation concealment: 4 low, 6 unclear, 1 high Blinding (participants and personnel): 11 high Blinding (outcome assessors): 2 low, 9 unclear Incomplete outcome data: 5 low, 5 unclear, 1 high Selective reporting: 2 low, 8 unclear, 1 high Other: 6 low, 5 unclear Overall: "Overall, the risk of bias was judged to be unclear to moderate."
Wojcieszek 2018	Sequence generation: 9 low, 1 high Allocation concealment: 8 low, 1 unclear, 1 high Blinding (participants and personnel): 4 low, 1 unclear, 5 high Blinding (outcome assessors): 10 low Incomplete outcome data: 7 low, 1 unclear, 2 high Selective reporting: 1 low, 9 unclear Other: 7 low, 3 unclear Overall: "We judged the risk of bias in the trials for methodology and reporting to be low to moderate."

Open in a new tab

Methodological quality of included reviews

When assessed against the ROBIS domains, 10 reviews (Bain 2014; Crawford 2015; Dodd 2018; Han 2012; Middleton 2018:Palacios 2019; Shepherd 2017; Spencer 2015; Tieu 2017; Wojcieszek 2018) were considered at low risk of bias across all domains, 'Study eligibility criteria', 'Identification and selection of studies', 'Data collection and study appraisal' and 'Synthesis and findings'. One review (Barrett 2014) was assessed as being at unclear risk of bias in the domains 'Data collection and study appraisal' and 'Synthesis and findings'. This review included six papers relating to one trial that was considered to have some design flaws. Only 45% of women, who were considered "high risk" for GDM were tested for GDM. The intervention group (probiotic with dietary intervention) numbers were halved in the Barrett 2014 review analyses as they performed two comparisons: "probiotics versus placebo" and "probiotics versus diet". The review comparison labelled "probiotics vs diet" refers to comparison of probiotic with dietary intervention versus placebo with dietary intervention with both these groups receiving dietary intervention. The review comparison labelled "probiotics versus placebo" refers to probiotics with dietary intervention versus placebo with no dietary intervention. As the GRADE approach was not used in the review to assess the quality of the evidence, we reviewed original trial reports in order assess their risk of bias and quality of evidence using GRADE. The trial included in Barrett 2014 was also in Tieu 2017, but the 'Risk of bias' assessment differed, which could be because different groups in the trial were being compared in these reviews. Barrett 2014 assessed the trial as low risk of bias across all domains. Tieu 2017 assessed risk of bias for the same trial as follows. Sequence generation: low; Allocation concealment: low; Blinding (participants and personnel): high, Blinding (outcome assessors): unclear; Incomplete outcome data: low; Selective reporting: unclear; Other: low.

Details of the ROBIS assessments we made can be found in Table 3.

3. ROBIS assessments for included reviews.

Review ID	ROBIS domains				Overall risk of bias
Review ID	Study eligibility criteria	Identification and selection of studies	Data collection and study appraisal	Synthesis and findings	Overall risk of bias
Bain 2014	Low risk	Low risk	Low risk	Low risk	LOW RISK
Barrett 2014	Low risk	Low risk	Unclear risk	Unclear risk	UNCLEAR RISK
Crawford 2015	Low risk	Low risk	Low risk	Low risk	LOW RISK
Palacios 2019	Low risk	Low risk	Low risk	Low risk	LOW RISK
Dodd 2018	Low risk	Low risk	Low risk	Low risk	LOW RISK
Han 2012	Low risk	Low risk	Low risk	Low risk	LOW RISK
Middleton 2018	Low risk	Low risk	Low risk	Low risk	LOW RISK
Shepherd 2017	Low risk	Low risk	Low risk	Low risk	LOW RISK
Spencer 2015	Low risk	Low risk	Low risk	Low risk	LOW RISK
Tieu 2017	Low risk	Low risk	Low risk	Low risk	LOW RISK
Wojcieszek 2018	Low risk	Low risk	Low risk	Low risk	LOW RISK

Open in a new tab

Effect of interventions

We have summarised the main results below. The outcome, GDM is presented for the different intervention types and categorised according to the framework described under 'Data synthesis': clear evidence of benefit; clear evidence of harm; clear evidence of no effect or equivalence; possible benefit; possible harm; or unknown benefit or harm. The first three categories represent GRADE moderate‐ or high‐quality evidence for which we found either clear benefit, clear harm or clear evidence of no effect (i.e. equivalence with a comparator). These categories are identified by a green tick, a red‐cross and a green equal‐sign icon, respectively (see Figure 1). For 'clear' benefit or harm, the confidence interval (CI) associated with the effect size did not cross the line of no effect. For 'clear evidence of no effect or equivalence' we considered a CI approximating the range of risk ratio (RR) 0.75 to 1.25 as sufficiently narrow to indicate a minimal effect relative to the comparator; these are thresholds recommended by GRADE (Guyatt 2011b). Please refer to Figure 3 for a summary of all assessments. Additional details can also be found in the 'Summary of findings' for each intervention type: Table 4; Table 5; Table 6; Table 7; Table 8; Table 9.

4. Summary of findings: dietary interventions.

Intervention and comparison	Outcome	Anticipated absolute effects (95% CI)		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
Intervention and comparison	Outcome	Assumed risk with comparator	Corresponding risk with intervention	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
Dietary advice vs standard care in pregnancy Tieu 2017	GDM	126 per 1000 82/651	76 per 1000 (44 to 131) 54/628	RR 0.60 (0.35, 1.04)	1279 (5)	Very low^a
Low GI vs mod‐high GI diet in pregnancy Tieu 2017	GDM	110 per 1000 49/444	100 per 1000 (70 to 145) 47/468	RR 0.91 (0.63, 1.31)	912 (4)	Low^b

Open in a new tab

a Studies contributing data had design limitations; There was considerable variation in the size of the effect in different studies; Wide 95% confidence interval crossing the line of no effect.

b Studies contributing data had design limitations; Wide 95% confidence interval crossing the line of no effect.

5. Summary of findings: exercise interventions.

Intervention and comparison	Outcome	Anticipated absolute effects (95% CI)		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
		Assumed risk with comparator	Corresponding risk with intervention
Any exercise vs routine care in pregnancy Han 2012	GDM	62 per 1000 24/389	68 per 1000 (41 to 114) 30/437	RR 1.10 (0.66, 1.84)	826 (3)	Low^a

Open in a new tab

a Study design limitations; wide 95% confidence interval crossing the line of no effect.

6. Summary of findings: combined diet and exercise interventions.

Intervention and comparison	Outcome	Anticipated absolute effects (95% CI)		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
		Assumed risk with comparator	Corresponding risk with intervention
Diet and exercise vs standard care in pregnancy Shepherd 2017	GDM	168 per 1000 551/3280	143 per 1000 (119 to 170) 525/3353	RR 0.85 (0.71, 1.01)	6633 (19)	Moderate^a

Open in a new tab

a Study design limitations: 19 RCTs, intervention unable to be blinded (not downgraded for this as outcome is objective); some RCTS with potentially serious design limitations (unclear randomisation, attrition bias); Inconsistency: I² = 42%, possibly largely due to one trial (Dodd 2014), not downgraded))

7. Summary of findings: dietary supplement interventions.

Intervention and comparison	Outcome	Anticipated absolute effects (95% CI)		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
Intervention and comparison	Outcome	Assumed risk with comparator	Corresponding risk with intervention	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
Omega‐3 fatty acid supplementation in pregnancy vs none Middleton 2018	GDM	62 per 1000 158/2569	63 per 1000 (51 to 77) 167/2666	RR 1.02 (0.83, 1.26)	5235 (12)	High
Myo‐inositol vs control in pregnancy Crawford 2015	GDM	283 per 1000 75/265	122 per 1000 (82 to 181) 27/237	RR 0.43 (0.29, 0.64)	502 (3)	Low^a
Probiotics and diet vs placebo and diet in pregnancy Barrett 2014	GDM	333 per 1000 27/76	131 per 1000 (53 to 316) 5/38	RR 0.37 (0.15, 0.89)	114 (1)	Very low^b
Probiotics and diet vs control in pregnancy Barrett 2014	GDM	342 per 1000 25/73	130 per 1000 (55 to 315) 5/38	RR 0.38 (0.16, 0.92)	111 (1)	Very low^b
Vitamin D vs control/placebo in pregnancy Palacios 2019	GDM	127 per 1000 25/197	65 per 1000 (34 to 123) 13/249	RR 0.51 (0.27, 0.97)	446 (4)	Low^c
Vitamin D+calcium vs control/placebo in pregnancy Palacios 2019	GDM	37 per 1000 1/27	12 per 1000 (0 to 290) 0/27	RR 0.33 (0.01, 7.84)	54 (1)	Very low^d
Vitamin D+calcium+other vitamins or minerals vs calcium+other vitamins or minerals in pregnancy Palacios 2019	GDM	12 per 1000 3/259	5 per 1000 (1 to 20) 5/1039	RR 0.42 (0.10, 1.73)	1298 (1)	Very low^e

Open in a new tab

a Unclear risk of bias for allocation concealment in two of the included trials (one trial did not provide sufficient detail to determine allocation concealment and one trial (reported as a conference abstract) had no details of random sequence generation, allocation concealment or blinding) and for high risk of performance bias for lack of blinding (two trials were open‐label trials with no blinding of participants or researchers, however one trial explicitly described blinding of outcome assessors and was assessed as low risk of detection bias); Studies were conducted in Italy with Caucasian women and generalisability of findings is limited.

b Serious limitations in study design as not all women tested for GDM. Very serious imprecision with one small study and a wide 95% confidence interval crossing the line of no effect.

c Serious limitations in study design due to one study being assessed as high risk of bias for several domains. Serious limitations due to indirectness with studies performed only in Asian countries (this downgrading was added by overview authors).

d Serious limitations in study design due to one study being at high risk of performance and detection bias; Very serious limitations in imprecision due to one small study, with a single event and wide 95% confidence intervals crossing the line of no effect contributing data.

e Very serious limitations in imprecision with only one study, with few events, and wide 95% confidence intervals crossing the line of no effect; Serious indirectness as there were multiple nutrient interventions in addition to vitamin D.

8. Summary of findings: pharmaceutical interventions.

Intervention and comparison	Outcome	Anticipated absolute effects (95% CI)		Relative effect RR (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
Intervention and comparison	Outcome	Assumed risk with comparator	Corresponding risk with intervention	Relative effect RR (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
Metformin vs placebo in pregnancy Dodd 2018	GDM	143 per 1000 64/448	121 per 1000 (87 to 170) 53/444	0.85 (0.61, 1.19)	892 (3)	Moderate^a
Care prior to and during subsequent pregnancies following stillbirth. Low molecular weight heparin versus no treatment/standard care. Wojcieszek 2018	GDM	125 per 1000 5/40	160 per 1000 (63 to 406) 8/45	1.28 (0.50, 3.25)	85 (2)	Low^b
Care prior to and during subsequent pregnancies following stillbirth. Low‐dose aspirin versus placebo. Wojcieszek 2018	GDM	182 per 1000 2/11	76 per 1000 (7 to 738) 1/13	RR 0.42 (0.04, 4.06)	24 (1)	Very low^c
Care prior to and during subsequent pregnancies following stillbirth. Low‐dose aspirin + low molecular weight heparin versus low‐dose aspirin. Wojcieszek 2018	GDM	77 per 1000 1/13 (aspirin)	21 per 1000 (1 to 479) 0/16	0.27 (0.01, 6.23)	29 (1)	Very low^c
Care prior to and during subsequent pregnancies following stillbirth. Low‐dose aspirin + low molecular weight heparin versus placebo Wojcieszek 2018	GDM	182 per 1000 2/11	25 per 1000 (2 to 487) 0/16	0.14 (0.01, 2.68)	27 (1)	Very low^c
Care prior to and during subsequent pregnancies following stillbirth. Low molecular weight heparin versus low‐dose aspirin. Wojcieszek 2018	GDM	0 per 1000 0/12 (aspirin)	0 per 1000 (not estimable) 0/10	no events	22 (1)	Very low^d
Care prior to and during subsequent pregnancies following stillbirth. Low molecular weight heparin (dose adjusted according to anti‐factor Xa levels) versus low molecular weight heparin (fixed dose). Wojcieszek 2018	GDM	143 per 1000 1/7 (fixed dose)	54 per 1000 (3 to 1000) 0/6	0.38 (0.02, 7.93)	13 (1)	Very low^e
Care prior to and during subsequent pregnancies following stillbirth. Leukocyte immunisation versus placebo. Wojcieszek 2018	GDM	0 per 1000 0/2	0 per 1000 (not estimable) 0/2	no events	4 (1)	Very low^f
Care prior to and during subsequent pregnancies following stillbirth. Intravenous IgG versus placebo. Wojcieszek 2018	GDM	0 per 1000 0/4	0 per 1000 (not estimable) 0/3	no events	7 (2)	Very low^f

Open in a new tab

a Study design limitations as one study of the three studies included has unclear risk of bias for random sequence generation, allocation concealment, performance bias, outcome assessor bias and selective reporting bias and reports a much greater effect than the other two studies.

b Limitations in study design; wide 95% confidence intervals crossing the line of no effect.

c Serious limitations in study design; very serious limitations in imprecision with only one study, with few events, and wide 95% confidence intervals crossing the line of no effect.

d Serious limitations in study design; very serious limitations in imprecision with only one study, with few events, and wide 95% confidence intervals crossing the line of no effect.

e Limitations due to indirectness as study population at increased risk of pregnancy loss and GDM not a predefined outcome; Very serious limitations in imprecision with only one study, with few events, and wide 95% confidence intervals crossing the line of no effect.

f Limitations due to indirectness as study population at increased risk of pregnancy loss and GDM not a predefined outcome; Very serious limitations in imprecision with only one study, with few events, and wide 95% confidence intervals crossing the line of no effect.

9. Summary of findings: management of other health issues in pregnancy.

Intervention and comparison	Outcome	Anticipated absolute effects (95% CI)		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
Intervention and comparison	Outcome	Assumed risk with comparator	Corresponding risk with intervention	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
Universal screening for thyroid dysfunction vs risk based screening in pregnancy Spencer 2015	GDM	40 per 1000 90/2257	37 per 1000 (28 to 50) 84/2259	RR 0.93 (0.70, 1.25)	4516 (1)	Moderate^a
Fractional Exhaled Nitrogen Oxide algorithm vs clinical guideline algorithm to adjust asthma therapy in pregnancy Bain 2014	GDM	104 per 1000 11/106	77 per 1000 (32 to 184) 8/104	RR 0.74 (0.31, 1.77)	210 (1)	Low^b
Pharmacist led multidisciplinary approach to management of maternal asthma vs standard care in pregnancy Bain 2014	GDM	69 per 1000 0/29	345 per 1000 (17 to 1000) 2/29	RR 5.00 (0.25, 99.82)	58 (1)	Low^b

Open in a new tab

a Serious limitation in imprecision with a wide confidence interval crossing the line of no effect

b Very serious limitation in imprecision as a single study with a small sample and a wide 95% confidence interval crossing the line of no effect.

Interventions for prevention of GDM