The effectiveness of pharmacological and lifestyle interventions to reduce the risk of diabetes and hyperglycaemia following gestational diabetes: A systematic review and meta‐analysis

To synthesize the available evidence to better understand the effectiveness of interventions to prevent or delay hyperglycaemia and Type 2 diabetes mellitus (T2DM) postnatally in women with current or previous gestational diabetes mellitus (GDM).


| INTRODUCTION
Gestational diabetes mellitus (GDM) is defined as raised blood glucose levels that appear in pregnancy, and usually resolves following delivery. 1,2There are multiple diagnostic criteria, such as those from the International Association of Diabetes and Pregnancy Study Groups that diagnose GDM through an oral glucose tolerance test (OGTT) when fasting plasma glucose is ≥5.1 mmol/L, 1-h plasma glucose ≥10.0 mmol/L and/or 2-h plasma glucose ≥8.5 mmol/L. 3lthough prevalence estimates vary between and within regions, GDM is common, affecting nearly one in seventeen pregnancies in Europe and one in eight pregnancies in the Middle East and North Africa. 4 history of GDM represents the single largest risk factor for the development of Type 2 diabetes mellitus (T2DM) in affected women.6][7] Risk of progression to T2DM may be greatest during the first 5 years after pregnancy, 8 but persists over the life course. 7Approximately a third of women with GDM are diagnosed with T2DM by 15 years postpartum. 7Factors such as elevated body mass index (BMI), non-White European ethnicity, particularly Asian ethnicity, and poorer pregnancy glucose tolerance that required treatment with insulin have been found to further increase risk of T2DM. 7,9,10omen with GDM are encouraged to maintain healthy lifestyles during and after pregnancy, in particular to establish a healthy BMI through being physically active and eating healthily.][13] Two previous reviews have assessed the effectiveness of pharmacological interventions such as metformin and have found some evidence that they prevent T2DM in women with GDM. 14,15However, there was a paucity of studies included and since both reviews were published in 2014, data from more recent studies are now available for consideration.Several reviews have synthesised data on the effectiveness of lifestyle or behavioural interventions.Of note, Gilinksy synthesised 13 studies and found little evidence for the benefit of lifestyle interventions in reducing T2DM risk (considering behavioural, anthropometric and glycaemic/T2DM incidence outcomes), 16 as did Miyazaki's overview of reviews. 17 reported that lifestyle interventions during the postpartum period (within 3 years of GDM) reduced T2DM risk, whereas interventions during pregnancy did not. 18edersen's narrative synthesis suggested that interventions that began after the first 6 weeks following delivery might be more effective than interventions that started any earlier. 19Overall, this suggests that intervention effects may vary according to factors such as setting, format and timing relative to pregnancy with many questions outstanding.
We aimed to synthesize the available evidence to reduce uncertainty concerning the effectiveness of interventions at preventing or delaying the progression to T2DM in women with current GDM or a history of GDM.We considered both pharmacological and lifestyle interventions that could begin during or after pregnancy affected by GDM, and included studies with both T2DM and continuous glycaemic outcomes.

| METHODS
The protocol for this systematic review was registered on PROSPERO (https:// www.crd.york.ac.uk/ prospero) diabetes mellitus, Type 2, diabetes, gestational, life style, meta-analysis, risk management, systematic review

What's new?
• There is uncertainty concerning the effectiveness of interventions at preventing or delaying the progression to Type 2 diabetes in women with gestational diabetes.

| Search strategy
We searched MEDLINE, Embase, PsycINFO, CINAHL and the Cochrane Library from inception up to September 2017, and later updated the search in December 2020, using the search strategies shown in Table S1.No language or other restrictions were applied.

| Study selection
We included articles that reported primary data published in peer-reviewed journals.The incidence of T2DM, or other glycaemic measures must have been quantified in any women with previous GDM following an intervention.This could be a pharmacological or lifestyle/behavioural intervention that aimed to improve postpartum glycaemic or cardiometabolic outcomes or health behaviours among mothers, distinct from usual or routine GDM pregnancy care.The intervention could commence antenatally but glycaemic or T2DM incidence outcomes must have been measured postpartum.The diagnostic method or criteria for both GDM and T2DM (if reported) must have been specified.All study designs were eligible.Amongst the titles and abstracts identified during the literature search, duplicate citations were removed.These were then assessed by two authors against the study's inclusion and exclusion criteria.Studies not relevant were rejected at this stage.All authors involved reviewed 10% of the papers independently to assess for any discrepancies between authors' decisions.Any discrepancies in opinion were reviewed and clarified by the remaining co-authors.Rayyan, an online systematic review support tool, was used to manage citations and authors' inclusion/exclusion decisions for the titles and abstracts from the updated search. 20Finally, full articles were reviewed by at least two authors independently according to the same criteria.Any discrepancies in opinion were again discussed and reviewed by the remaining co-authors.

| Data extraction
A data extraction form was developed by RW in Microsoft Excel and enhanced after discussion with the co-authors.This was used to facilitate systematic extraction of summary study, incidence and demographic information from the included citations.We tested the data extraction form in a randomly selected sample of papers from the initial literature search to ensure it was suitable for gathering the required data to answer the research question.Data were extracted by two authors independently and any differences were resolved by discussion.
We classified studies into three groups based on timing that the intervention was implemented (pregnancy, pregnancy and postpartum, postpartum).Extracted data included study location, study design, study period and recruitment context.We determined whether glycaemic tests, medical records or self report had been used to identify GDM or T2DM.Additionally, sample sizes, the nature of interventions and follow-up timepoints were extracted.Finally, we extracted study-level demographic characteristics, including age (mean, median or number in specific age groups), ethnicity, gestational age at delivery, BMI and smoking status.
If data from the same study population had been reported in multiple publications, we included the publication with the longest follow-up or the main publication if the same data were reported again.

| Risk of bias
We then assessed the quality of each included study using the Critical Appraisal Skills Programme (CASP) checklist for randomised controlled trials (RCTs). 21The response set for nine of the eleven questions was 'yes', 'no' or 'can't tell', for one question it was 'large', 'small' or 'no' effect size and for one question it was 'significant', 'insignificant' or 'no' effect.Quality assessment was independently completed by two authors for each study.
We considered that seven of the questions in the CASP checklist assessed risk of bias.Studies that were marked as 'yes' for three or fewer of these seven questions were considered to have a high risk of bias, studies that were marked as 'yes' for four or five questions were considered to have a moderate risk of bias and studies that were marked as 'yes' for six or seven questions were considered to have a low risk of bias.

| Data synthesis
To conduct a meta-analysis, a minimum of two studies evaluating the same intervention-outcome combination were needed, therefore only pharmacological and physical activity and/or diet interventions were included in the meta-analyses.Furthermore, to be included the intervention had to be compared with a control arm.The results of the other studies were included in the narrative review only.
Continuous data were harmonised in order to pool results.For each measure, we converted results into the same units.Median and interquartile range (IQR) were converted to mean and standard deviation (SD). 22We estimated the standardised mean difference (SMD) and 95% confidence intervals (CIs) for continuous measures using Hedges' g to account for small sample sizes. 23or the binary outcome, incidence of T2DM, we used count data to calculate relative risk (RR) and 95% CIs.Two studies each with two pharmacological intervention arms reported T2DM outcomes. 24,25Within each study, the results from the two arms were combined to provide an overall estimate of the effect of pharmacological interventions.In one study 26 there were no T2DM outcomes at follow-up; in order to include this study in meta-analyses, we increased the number of events and individuals in each arm by one. 27harmacological interventions were stratified by timing of intervention as the effects are likely to be transient, and lifestyle interventions were stratified by intervention type as these may change longer term habits.When pooling results, we used the inverse variance weighted random-effects meta-analysis to produce a pooled RR.We used DerSimonian-Laird models to allow for betweenstudy heterogeneity as clear differences between studies were identified, such as for participant ethnicity.We assessed heterogeneity using the Cochrane chi-squared statistic and the I-squared statistic.Forest plots displayed individual RR for T2DM or SMD for continuous measures, as well as summary estimates.
To evaluate whether associations varied by levels of a third variable, we conducted meta-regression analyses.A minimum of 10 studies were required for meta-regression, therefore only studies assessing the impact of diet and physical activity interventions on fasting plasma glucose or T2DM incidence were included.The impact of the following study-level variables was assessed: timing of intervention (pregnancy, postpartum or both), duration of intervention, time from study onset to assessment, BMI at baseline, age at baseline and percentage of participants who were of White ethnicity.
We assessed potential publication bias through funnel plots and Egger tests for meta-analyses including at least 10 studies. 28e conducted two sensitivity analyses.To assess the impact of study bias, we excluded studies that were at a high risk of bias from meta-analyses and reproduced forest plots.As the DerSimonian-Laird method for random effects meta-analysis may have statistical limitations in the case of few studies, 29 we reran all meta-analyses with less than 10 studies with fixed-effects models to assess the consistency of the results and provide an overall estimation of the relationship specifically in the populations studied.
All tests were two-tailed and p-values of <0.05 were considered statistically significant.We used Stata software package (version 15; Stata Corp, College Station, TX) for all statistical analyses.

| Study selection
The original search identified 23,160 distinct citations, with a further 5239 citations arising from the second search (Figure 1).We reviewed 355 full texts for eligibility and included 31 articles in the systematic review.The most common reason for exclusion was citations found not to be peer-reviewed articles reporting new, original data (e.g.posters, protocols and systematic reviews; n = 169 articles), followed by no measure of glycaemia postpartum (n = 53 articles).
Five studies did not include a placebo or usual care control group therefore were not included in the meta-analyses.

| Study characteristics
The characteristics of 31 studies included in the review are summarised in Table 1.Study characteristics are reported in detail in Table S2.All of the studies except Kapoor et al. 30 were RCTs.The studies took place between 1989 and 2019.Sixteen different countries were represented, with most studies based in the USA (n = 10 studies) or Australia (n = 6 studies).In total, 8624 participants with GDM or a history of GDM were included at baseline and 6144 participants were followed up (excluding two studies in which sample size at follow-up was not clearly reported).Median baseline sample size was 113 participants and ranged from 21 to 2280 participants for individual studies.Most studies (n = 22) followed up participants at 1 year or less, six studies followed up within 3 years, and three studies followed up within 3-10 years.
Twenty-one studies reported T2DM outcomes.The median annualised incidence of T2DM in the intervention arms of these studies was 5.1 cases per 100 women-years (IQR 1.2-6.8),and 6.3 cases of T2DM per 100 women-years (IQR 4.2-10.3) in the control arms (Figure 2).Twenty-two studies reported continuous measures of glycaemia: fasting plasma glucose, 2-h OGTT and/or HbA1c.This includes both absolute measures of glycaemia at follow-up from which we calculated SMD (n = 14 studies), and/or change in measures of glycaemia between baseline and follow-up only (n = 11 studies).
Table S3 summarises the characteristics of the participants.Average age at the time of the study ranged from 26 to 43 years.Most participants were overweight according to the mean BMI or weight classifications reported for the studies.Ten studies included participants of a single ethnicity.
The interventions are described and classified in Table S4.Six interventions took place during pregnancy only, two started during pregnancy and continued after delivery and the remaining twenty-three interventions took place solely postpartum.The duration of these interventions ranged from 4 weeks to up to 10 years for the Diabetes Prevention Program Outcomes Study, 24 although only 10 lasted for a year or longer.There were nine studies including pharmacological interventions (six studies investigated various glucose-lowering medications, two studies investigated Vitamin D supplementation and one study investigated oral contraceptives) and 21 studies included non-pharmacological interventions (three studies were diet-only, two studies were physical activity-only and eighteen studies involved both diet and physical activity).Fifteen of the dietary and physical activity interventions included individual and/or group face-to-face elements, and fifteen included remote support such as online resources, smartphone applications and text messages.Typically, the control groups received standard care that included generic post-GDM advice or information.

| Risk of bias
Our risk of bias assessment identified six studies with a low risk of bias, twenty-one studies with a moderate risk of bias and four studies with a high risk of bias (Table S5).The most frequent source of bias was from not blinding to study arm.Whilst it was not possible to blind participants to lifestyle intervention arms, study statisticians and those taking measurements or processing samples tended not to be reported to be blinded either.Attrition bias was similarly common, with high loss to follow-up reported.
Furthermore, we did not identify a risk of publication bias (p > 0.05, Egger's test).Funnel plots were only produced for the meta-analyses with 10 or more studies included (Figure S1).

| Effect of intervention on the risk of T2DM
When the studies were pooled, both pharmacological and lifestyle interventions reduced the incidence of T2DM diagnosed in women with a history of GDM (Figure 3).The greatest risk reduction was seen for pharmacological interventions (RR 0.80 [95% CI 0.64-0.999],n = 6 studies), with a larger risk reduction when the pregnancy intervention was excluded (RR 0.79 [95% CI 0.66-0.95],n = 5 studies).Overall, physical activity and/or diet interventions Excluding studies with a high risk of bias increased the magnitude of the RR estimates but without statistical significance (Figure S2).In the fixed-effects meta-analysis, the pooled effectiveness of both pregnancy and postpartum pharmacological interventions was statistically significant (Figure S3).

| Effect of intervention on continuous glycaemic measures
When considering the SMD of continuous measures of glycaemia -fasting plasma glucose, 2-h OGTT and HbA 1cseparately, no positive associations with pharmacological interventions were observed (Figure 4a).In fact, placebo arms were favoured, although this was not statistically significant.As only two studies were pooled and Valizadeh et al. 31 had a high risk of bias, only one study remained in the sensitivity analysis (Figure S4a).No differences were observed between the random and fixed-effects meta-analyses (Figure S5a).Two studies reported change in fasting plasma glucose and HbA 1c between baseline and follow-up with pharmacological interventions (Figure S6a).When these were pooled, a decrease in both measures was observed but this was not statistically significant (change −2.17 [−5.97-1.63]for fasting plasma glucose and − 0.14 [−0.73-0.44]for HbA 1c ).
Conversely, dietary and physical activity interventions were found to be associated with a small reduction in participants' fasting plasma glucose (SMD −0.23 [95% CI −0.49 to 0.04], n = 12 studies) (Figure 4b).This was not statistically significant in the main analysis, nor when one study with a high risk of bias was excluded in the sensitivity analysis (SMD −0.26 [95% CI −0.54 to 0.01], n = 11 studies) (Figure S4b).Similar to the pharmacological interventions, however, there were small increases in the SMD of 2-h OGTT and, particularly, HbA 1c when eight and four studies were pooled, respectively.Excluding studies with a high risk of bias-one study reporting 2-h OGTT and none reporting HbA1c-had minimal impact on the effect sizes of the pooled SMDs (Figure S4b), as did conducting a fixed-effects meta-analysis (Figure S5b).Considering the The annualised incidence of T2DM of studies included in this review, grouped by intervention type.* Indicates where the same annualised incidence was observed for multiple study arms.Elkind-Hind 2020a annualised incidence in Intervention 1 was 0.83 per 100 women-years and Intervention 2 was 0.
studies that only reported change in glycaemic outcomes with lifestyle interventions (Figure S6b), no change in fasting plasma glucose (n = studies) or HbA 1c (n = 3 studies) was observed, but 2-h OGTT decreased by 0.11 units (95% CI −0.23 to 0.00, n = 5 studies).intervention (pregnancy and/or postpartum), and longer durations of interventions and intervals between the intervention and assessment of T2DM status were not statistically significantly associated with a reduction in risk of T2DM.Similarly, the intervention effect did not clearly vary according to average maternal BMI, age or ethnicity (percentage White ethnicity) at the study level.

| Association of study characteristics with continuous glycaemic measures
The timing of the intervention, duration of time to assessment or average maternal characteristics did not alter the association between diet and/or physical activity and fasting plasma glucose (Table 3).However, longer interventions were associated with a greater average decrease in fasting plasma glucose-for each additional month of intervention, SMD of fasting plasma glucose decreased by 0.009 units (95% CI −0.016 to −0.001, p = 0.024).In the pharmacological interventions, metformin was compared to other glucose-lowering medications (where metformin-only acted as the comparator arm in these studies).Daniele et al. 32 found that sitagliptin combined with metformin was associated with reduced fasting plasma glucose at 16 weeks follow-up compared to baseline (change −0.49 ± 0.01 mmol/L, p < 0.05), while metformin or sitagliptin alone were not associated with statistically significant changes.T2DM was diagnosed in none of the 40 participants in that study.In two studies by Elkind-Hirsch et al., 33,34 liraglutide or dapagliflozin plus metformin were associated with positive effects on glycaemic outcomes in women with recent GDM.While fasting blood glucose decreased in all arms after 24 weeks of treatment, mean blood glucose during OGTT decreased significantly more in the dapagliflozin-only and dapagliflozin plus metformin arms (change −11 mg/dL and −10 mg/dL respectively, p = 0.05 for metformin versus dapagliflozin and dapagliflozin plus metformin). 33Similarly, liraglutide plus metformin was associated with a greater decrease in fasting blood glucose levels and mean glucose levels during OGTT after 84 weeks of treatment compared to the metformin-only arm, although the overall effect of treatment was only statistically significant for mean glucose levels during OGTT. 34 the lifestyle interventions, two further studies reported conversion to T2DM in study participants.In Kapoor et al. 30 the incidence of T2DM was 1.8% (n = 56) in the single-arm evaluation of the 6-month group lifestyle programme.In Shyam et al. 35 it was 8.0% (n = 25) in the personalised conventional healthy diet recommendations arm and 3.7% (n = 27) in the personalised low glycaemic diet recommendations arm.Kapoor et al. 30 further reported that compared to before the intervention, average fasting plasma glucose decreased by 0.3 mmol/L (p = 0.03) and 2-h OGTT decreased by 0.9 mmol/L (p < 0.001) at 6 months follow-up.Shyam et al. 35 reported slightly smaller improvements in their six-month study, where median fasting plasma glucose and 2-h OGTT both decreased by 0.2 mmol/L in the personalised low glycaemic diet recommendations arm.

| DISCUSSION
In this systematic review, we found that interventions in women with a previous history of GDM were beneficial for reducing the incidence of T2DM after pregnancy, although the effects were modest.Postpartum pharmacological interventions were most effective, particularly glucose-lowering medication.Considering individual measures of glycaemia separately for lifestyle interventions suggested that the greatest benefits of interventions were seen for fasting plasma glucose rather than 2-h OGTT and HbA 1c outcomes.Longer diet and/ or physical activity interventions were significantly more effective at improving fasting plasma glucose over the duration studied.
We observed an apparent discrepancy between the pooled effect of lifestyle intervention on T2DM and fasting plasma glucose, and 2-h OGTT and/or HbA 1c .This trend was observed within some of the individual studies such as those led by Cheung et al., Wein et al. and Juliusdottir et al. [36][37][38] However, the significance of these differences is likely due to random error and the relatively small number of studies that we included with these outcomes and the small number of participants in each of these studies.How outcomes were ascertained and completeness of follow-up could also have contributed.Alternatively, it is possible that the interventions had a greater impact on fasting plasma glucose than other outcomes.It is not possible to tell whether the diagnosis of T2DM was based on the fasting or 2-h OGTT outcome (since in many definitions, only one of the measurements needs to be above the cut-off for a diagnosis of T2DM), therefore the findings could be consistent and only fasting plasma glucose improved.Additional differences between meta-analyses using change in continuous glucose measures and continuous glucose outcomes at follow-up further highlight the ongoing uncertainty.
Unsurprisingly, pharmacological interventions had the most significant impact on T2DM incidence in the meta-analysis, and metformin plus an additional glucose-lowering agent was found to be the most effective combination in the narrative synthesis.The only pharmacological interventions to report continuous measures of glycaemia at follow-up were those that investigated the effects of vitamin D, therefore we were not able to meta-analyse the impact of glucose-lowering agents on fasting plasma glucose, 2-h OGTT or HbA 1c .While it is unclear whether the glucose-lowering agents actually prevented T2DM or effectively treated/managed it prior to progression and diagnosis (particularly as only two of the eight postpartum pharmacological interventions clearly describe a washout period between the end of the intervention and outcome assessment), they may provide an acceptable and feasible approach to management of hyperglycaemia after GDM for some women.
Unlike pharmacological interventions that cease when the intervention ends (that is, treatment is removed), lifestyle interventions may lead to changes in diet and physical activity that can be maintained after the study.Although the postpartum period may be a more appropriate time to learn and establish new routines, 12,39 and previous reviews have highlighted greater effectiveness for interventions that start after delivery, 18,19 this may explain why we did not observe significant differences according to timing relative to pregnancy.Prompted by recommendations based on a qualitative synthesis, we have previously identified a wide range of intervention elements that may help to support women to reduce their T2DM risk by making changes to their diet and physical activity. 12Practical advice and individualised information, monitoring and feedback, peer support and remote support were all anticipated to be useful by some participants and at various stages.Together with the findings of this review, this supports the hypothesis that there is no perfect, universal approach to preventing T2DM in this population, but that it is worthwhile to implement and refine such interventions nonetheless.
Although still potentially effective, the magnitude of the benefits of lifestyle interventions that we found were smaller and less statistically significant than that of other reviews.Specifically, Li et al.'s meta-analysis of lifestyle interventions found that interventions that started within three years of the index GDM pregnancy were highly effective in reducing the risk of postpartum T2DM with a pooled fixed-effects RR of 0.57 (95% CI 0.42-0.78,n = 10 studies). 18This may be explained by the different studies they included due to different criteria and the inclusion of Chinese-language articles and those identified through Chinese databases.Previous reviews of pharmacological interventions have not conducted meta-analyses. 14,15ther reviews have reported mixed conclusions on this subject.Miyazaki et al. 2017 conducted an overview of reviews of non-pharmacological interventions to prevent T2DM after GDM. 17 They concluded that there was insufficient evidence that lifestyle interventions lower the risk of T2DM nor improve glycaemic load.Conversely, Hedeager Momsen et al. found that postpartum lifestyle interventions were effective in slightly decreasing diabetes incidence in their overview of reviews, although they also considered breastfeedingonly interventions following GDM and did not conduct a meta-analysis. 13n this systematic review, we have updated the literature on the pooled effectiveness of both pharmacological and lifestyle interventions.A strength of the review is that we have included the key outcome of progression to T2DM as well as continuous measures of glycaemia that facilitate a more comprehensive understanding.Furthermore, we did not limit the inclusion criteria to studies with a placebo or usual care control arm.Because we needed to rely on the description of the studies included within the articles, it could be argued that certain studies should have been included/excluded or classified differently in the analyses.We followed the recommended methods for systematic reviews and such decisions were considered by at least two authors, including those with clinical backgrounds and expertise in diabetes healthcare.Although we identified 31 studies for inclusion in the review and 25 studies for the meta-analyses, grouping these according to key intervention characteristics resulted in a low number of studies in some categories (particularly continuous outcomes for pharmacological interventions) and therefore there was a low power to identify pooled effects and other influences on effect sizes.The small number of pharmacological intervention studies overall and the inconsistent reporting of study and participant characteristics meant that we were only able to conduct the meta-regression for the non-pharmacological interventions.The studies were primarily conducted in high income countries, therefore such interventions could be more or less effective in other contexts where participants may have lower health literacy or where GDM was less well controlled in pregnancy.However, we did not observe significant differences in effectiveness according to maternal age, BMI or percentage of the study population of White ethnicity.Furthermore, many studies had relatively short follow-up periods and so it is unclear whether T2DM can be effectively prevented in the long-term, although delaying progression is still likely to be valuable in terms of long-term cardiometabolic health. 40n conclusion, we found some evidence that both pharmacological and lifestyle interventions reduce hyperglycaemia and delay progression to T2DM after GDM, albeit not always with statistical significance.Nonetheless, uncertainty remains, particularly over the most effective approaches, which measures of glycaemia they influence and the mechanism of action.
Conversely, other reviews have found interventions to be effective: Li et al.

F I G U R E 3
Relative risk of T2DM associated with (a) pharmacological interventions stratified by timing of intervention, and (b) lifestyle interventions stratified by intervention type.Plots are ordered by duration of intervention.CI, confidence interval; RR, relative risk.Timing of intervention: preg-during pregnancy, post-postpartum, both-pregnancy and postpartum.

3. 8 |
Narrative synthesis of studies excluded from the meta-analyses Five studies were not included in the meta-analyses: three pharmacological interventions did not include placebo (a) F I G U R E 4 Standardised mean difference of continuous glycaemic measures (fasting plasma glucose, 2-h OGTT and HbA 1c ) associated with (a) pharmacological interventions, and (b) lifestyle interventions.Plots are ordered by duration of intervention.CI, confidence interval; OGTT, oral glucose tolerance test; RR, relative risk.Timing of intervention: preg-during pregnancy, post-postpartum, both-pregnancy and postpartum.F I G U R E 4 (Continued) (b) arms, and two lifestyle interventions did not have control arms.All of these interventions were initiated postpartum.

and year Country Total sample size (baseline/follow-up) Pharmacological Lifestyle Intervention summary Intervention length/ duration Follow-up time point Risk of bias
gestational diabetes mellitus; T2DM, Type 2 diabetes mellitus.Summary of study characteristics, grouped by timing of intervention.
T A B L E 1

Country Total sample size (baseline/follow-up) Pharmacological Lifestyle Intervention summary Intervention length/ duration Follow-up time point Risk of bias
had a smaller pooled effect size without statistical significance (RR 0.88 [95% CI 0.76-1.01],n = 12 studies).

Table 2
shows that risk of T2DM in lifestyle interventions was not significantly associated with timing of the Associations of study variables and maternal characteristics with the risk of T2DM after GDM in lifestyle interventions.Associations of study variables and maternal characteristics with standardised mean difference of fasting plasma glucose after GDM in lifestyle interventions.
T A B L E 2Abbreviations: BMI, body mass index; CI, confidence interval.