Risk of type 2 diabetes mellitus after gestational diabetes mellitus: A systematic review & meta-analysis

Gestational diabetes mellitus (GDM) is defined as glucose intolerance first detected during pregnancy. The prevalence of GDM has increased more than 30 per cent over the past two decades1. As reported, the median prevalence of GDM globally ranges from 1.8 to 22.3 per cent2. GDM is associated with short and long-term adverse outcomes of both mothers and their respective offsprings, and is a well - known risk factor for developing type 2 diabetes mellitus (T2DM) after delivery. The rates of T2DM diagnosis after GDM range from two to 70 per cent, from six weeks to 28 yr postpartum3. Increasing prevalence of GDM and T2DM and their related complications lead to huge healthcare and economic costs45.

In light of these risks and the opportunity for preventive intervention, women with GDM are advised to have oral glucose tolerance test (OGTT) assessed at 6-12 wk postpartum6. However, studies reported that postpartum screening rates range from 13 to 82 per cent varying across geography, ethnicity and practice patterns. and is underused789. Furthermore, while there are various barriers of postpartum diabetes screening patient compliance with diabetes screening recommendations are inadequate10. Systematic review and meta-analysis previously showed that women with a history of GDM have a sevenfold risk of being diagnosed as T2DM than those without although the results of this study were synthesized despite heterogenous differences11. In the present study the relative risks (RRs) among all selected studies were included and sensitivity and subgroup analyses were conducted to identify the sources of the heterogeneity. Moreover, risks of being diagnosed as T2DM vary widely2, and therefore the disparities of T2DM diagnosis after GDM in different demographic subgroups to help health providers focus on the high-risk patient were assessed.

Material & Methods

Literature search and inclusion criteria: Twenty studies were hand-searched from the previous systematic review11 and did an electronic search of MEDLINE and Embase from January 1, 2009 to July 31, 2019 and did not apply any restrictions. The search of the Cochrane Library was from inception to July 31, 2019, without restrictions. Search terms were a combination of ‘gestational diabetes mellitus’, ‘pregnancy diabetes mellitus’, ‘diabetes, gestational’, ‘type 2 diabetes mellitus’, ‘diabetes mellitus, type 2’ and ‘non-insulin dependent diabetics mellitus’. In addition to the electronic search, reference lists and citations of relevant reviews and articles were hand-searched.

Prospective and retrospective cohort studies (PCS and RCS) in which women were diagnosed with GDM and normal blood glucose were searched for. The outcome was the diagnosis of T2DM at six weeks or later after delivery. The criteria of GDM and T2DM were not restricted. Studies of women with pre-existing diabetes mellitus were excluded.

Methodological quality assessment: The quality of included studies was assessed by a standardized checklist based on the Newcastle–Ottawa Scale (NOS)12. The NOS is a star rating system (0-9 stars) used for observational studies. For cohort studies, the criteria cover three domains: selection of participants, between-group comparability and ascertainment of outcome. Each item can get one star in selection and outcome domains and two stars in comparability domain if appropriate methods were reported1213.According to the final score, studies were classified as high (c7-9 stars), medium (5-6 stars) or low (0-4 stars) quality. Low quality (c7) study might reduce the credibility of results, so we excluded low quality studies in this meta-analysis.

Data abstraction: Participant and study characteristics and cumulative incidences of T2DM in the GDM and non-GDM groups were independently extracted by two authors using standardized tables. Disagreements were solved by discussion with the third author. If more than one report based on the same population was identified, the one with the most relevant and complete information was selected.

Statistical analysis: A Meta-analysis was carried out using Stata/MP (Version 14.0, StataCorp LLC, Texas, USA). Unadjusted, pooled relative risks (RRs) and 95 per cent confidence intervals (CIs) were calculated. Heterogeneity was assessed with Cochrane’s Q text and by calculating I² values. High heterogeneity was defined by either P≤0.10 or I² ≥60 per cent, median heterogeneity was defined by either P ≤0.10 or 30 per cent ≤ I²< 60 per cent and little or no heterogeneity was defined by either P>0.10 or I²< 30 per cent14. In cases of high heterogeneity, a random-effects model was used. Sensitivity analyses were conducted to identify the outliers by testing the outcome robustness after one study was removed. Subgroup analyses were performed to explore the sources of heterogeneity among studies by stratification according to mean maternal age, body mass index (BMI) at follow up, race/ethnicity, region, family history of diabetes mellitus, time interval of postpartum OGTT performed, GDM criteria, T2DM criteria and number of confounders matched. Begg’s test and Egger’s test were performed to investigate small sample bias and publication bias. A P<0.05 was considered statistically significant.

Results

Selection of studies: In total, 1957 records were identified through electronic database searching, 30 additional publications were identified through reference lists and 20 publications were included from a previous systematic review. Altogether,1809 titles and abstracts were screened after 198 duplicates were removed. Of 343 publications that were selected for full-text review, 304 were excluded for various reasons. Finally, 39 cohort studies involving 2,847,596 women were included in this meta-analysis. In these studies, 78,893 women were diagnosed as T2DM at six weeks or later after delivery (Fig. 1).

Close

Fig. 1

PRISMA flow diagram showing literature search.

Export to PPT

Characteristics of the studies: A total of 26 retrospective415161718192021222324252627282930313233343536373839 and 13 prospective cohort studies>5404142434445464748495051 conducted in different countries were considered for this meta-analysis. The participants varied widely in maternal age, BMI, family history of diabetes mellitus, ethnicity, length of follow up and time interval of postpartum OGTT performed. Moreover, diagnostic criteria of GDM and T2DM varied by country as well. In 15.4 per cent (6/39) of studies, the dropout rate was under 30 per cent. In 5.1 per cent (2/39) of studies, the dropout rate is between 30 and 60 per cent. In 38.5 per cent (15/39) of studies, none of the women dropped out. In 41.0 per cent (16/39) studies, the dropout rate was not recorded. In 76.9 per cent (30/39) of studies, women in two groups were matched by different confounders. In 23.1 per cent (9/39) of studies, confounders adjustment was not recorded (Table).

As per the NOS scores as shown in Fig. 2, 87 per cent (34/39) of studies included in this meta-analysis were of high quality, and 13 per cent (5/39) studies were of medium quality. The unadjusted RRs of women diagnosed as T2DM at six weeks or later after delivery ranged from 1.32 (95% CI, 0.46-3.37) to 47.25 (95% CI, 2.95-758.01), with a pooled unadjusted RR of 8.92 (95% CI, 7.84-10.14).The heterogeneity was defined as high with P<0.01, and I²=94.1 per cent (Fig. 3). Sensitivity analyses were conducted by recalculating the pooled RRs with included studies removed one by one. The results indicated that the pooled RRs were not affected by the exclusion of any individual study (Fig. 4).

Close

Fig. 2

Newcastle–Ottawa Scale scores of 39 included studies in meta-analysis.

Export to PPT

Close

Fig. 3

Forest plot of the risk of women diagnosed as type 2 diabetes mellitus (DM) after gestational DM. X-axis is plotted in log scale. Solid squares and horizontal lines indicate relative ratios and 95 per cent confidence intervals. The diamond represents the pooled relative risk (RR).

Export to PPT

Close

Fig. 4

Sensitivity analysis of women diagnosed as type 2 DM after gestational DM. Three vertical lines indicate the pooled RR and 95 per cent CI of all studies. Circles and horizontal dashed lines indicate recalculated RRs and 95 per cent CIs.

Export to PPT

Subgroup analyses indicated that maternal characteristics and the time interval of postpartum OGTT performed was associated with the RR of T2DM onset after GDM. Older maternal age and family history of diabetes mellitus increased the risk of T2DM after GDM. The incidence of T2DM after GDM is the highest within the first year after delivery. The RR of diagnosing T2DM after GDM is variable when studies were grouped according to race/ethnicity and geographic region. The RR of diagnosing T2DM after GDM was lower when more confounders were matched (Fig. 5).

Close

Fig. 5

Risk of women diagnosed as type 2 DM after gestational diabetes mellitus grouped by maternal characteristics, study characteristics and diagnostic criteria. The diamond represents the subtotal relative risk.

Export to PPT

These results suggest that race/ethnicity, region, family history and time interval of postpartum OGTT performed could explain the reason behind the heterogeneity among studies. However, mean maternal age, BMI at follow up, GDM criteria, T2DM criteria and number of confounders matched could not explain the same.

Publications bias: No apparent asymmetry was observed in the Begg’s funnel plot (Fig. 6) and Egger’s publication bias plot (Fig. 7). Results of the Begg’s test (P=0.200) and Egger’s test (P=0.380) were not significant.

Close

Fig. 6

Begg’s funnel plot of 39 publications.

Export to PPT

Close

Fig. 7

Egger’s publication bias plot of 39 publications.

Export to PPT

Discussion

This meta-analysis indicates that women with a history of GDM have near nine fold increased risk of being diagnosed as T2DM in the future compared with those without GDM. The magnitude of the association between GDM and T2DM suggests that more frequent assessment and effective interventions targeting eligible women are needed. American Diabetes Association and other professional organizations recommend diabetes screening at 6-12 wk postpartum for women with GDM5253. Despite the emphasis of multiple guidelines, the postpartum screening compliance rates are still typically low5455. In addition, from the present study it was evident that within the first year after delivery, the progression of T2DM increased steeply. So, healthcare providers should emphasize the importance of continuity in treatment and healthcare and women with GDM should attend the follow up programmes earlier and conduct OGTT at 6-12 wk postpartum. Furthermore, later long-time screening strategies and optimal screening frequency may be needed further studies to explore.

Maternal age, BMI, race/ethnicity and family history are associated with the prevalence of GDM and T2DM11. In this meta-analysis, the results of subgroup analyses corroborated that maternal age and family history of diabetes might be the risk factors for T2DM after GDM. Thus, older women or those with a family history should value antepartum counselling and postpartum diabetes screening more than other women with GDM.

It has been suggested previously that the prevalence of GDM varies with race/ethnicity225, with Asians and Hispanics reported to have a higher GDM prevalence than non-Hispanic Whites and Blacks5657. In the present study it was observed that Blacks and non-Hispanic Whites had a higher RR of developing T2DM after GDM than Hispanics and Asians, which was consistent with a large multi-ethnic cohort study25. Another study reported that Hispanics and Asians had the highest RR of T2DM after GDM44 however, the sample size was small and CIs were wide44. This inconsistency could be attributed to the sample size. Large multi-ethnic cohort studies are needed to verify that conjecture.

Besides race/ethnicity, regional disparity (geographic level) is an important influence factor of GDM prevalence. The Middle East and North Africa had the highest prevalence of GDM, followed by South-East Asia, Western Pacific, South America, Africa and North America, whereas Europe had the lowest prevalence2. Despite the relatively high prevalence, no eligible studies from North Africa or Africa were identified in our search, and only two studies from South-East Asia were included. The subgroup analysis indicated that the RR of T2DM after GDM in Europe and South-East Asia was higher than other geographic regions. Although the GDM prevalence in Europe was the lowest, the RR of T2DM after GDM in Europe was the highest. Moreover, RRs in South America and Middle East were relatively low. Taken together, the RR of T2DM after GDM was not associated with GDM prevalence.

In this meta-analysis (P<0.01, I²=94.1%) high heterogeneity was noted similar to a previous study10 (P<0.01, I²=85%). In this meta-analysis, sensitivity analysis indicated that no individual study contributed to the heterogeneity and the subgroup analyses, indicated that maternal age, BMI at follow up, GDM and T2DM criteria, and number of confounders matched could not explain the heterogeneity. Nevertheless, race/ethnicity, region, family history and time interval of postpartum OGTT performed might have contributed to the same. In subgroup analysis based on race/ethnicity, no significant evidence of heterogeneity was found in group ‘non-Hispanic White’ and ‘Asian’, but significant evidence of heterogeneity was found in group ‘Other’ and ‘Hispanic’. In group ‘Other’, most studies included mixed population and their racial/ethnic composition was different, which was considered the cause of the subgroup heterogeneity. In group ‘Hispanic’, two studies were carried out in Europe and one in South America; it was thus inferred that regional disparity might cause subgroup heterogeneity. In the results of subgroup analyses based on geographic regions, we only observed significant evidence of heterogeneity in the group ‘North America’. Such heterogeneity might be attributed to diversity in race/ethnicity, because the degree of diversification among population in North America was higher than that among the population of other geographic regions and most studies on this group included mixed population. In subgroup analysis based on family history, no heterogeneity was found in the group ‘<25 per cent’ and ‘>25 per cent’. In addition, in subgroup analysis based on time interval of postpartum OGTT performed, no heterogeneity was found in the groups ‘at six weeks’ and ‘<one year’ and high heterogeneity was seen in group ‘>one year’. Therefore, it was inferred that the family history of diabetes and time interval of postpartum OGTT performed might be the source of heterogeneity. Meanwhile, 76.9 per cent (30/39) studies did not record the family history information and 41.0 per cent (16/39) studies did not record the time interval of postpartum OGTT performed. Such absence of information might have caused a bias.

There were, however, two limitations in the present study. The RR was synthesized regardless of the huge variance in diagnostic criteria and screening protocol for GDM and T2DM. However, the diagnose criteria have been constantly changing over the last four decades. In 1997, the T2DM diagnosis threshold was reduced58. Moreover, recent studies using the new International Association of Diabetes and Pregnancy Study Group criteria show a higher prevalence of GDM58. Therefore, the inclusion of old studies might have caused the underestimation of the risk of having T2DM after GDM. Secondly, the main source of heterogeneity in this study could not be identified. Such heterogeneity in the present study might have been caused by the number of included studies and the differences in the participant characteristics.

In summary, the high risk of diagnosing T2DM after GDM suggests that healthcare providers need postpartum screening and follow up programmes, both of which are convenient and economic methods for early treatment of T2DM, thereby reducing the prematurity of cardiovascular, renal and retinal diseases59606162. Continuous assessment and effective interventions targeting eligible women are needed, in particular, older women with GDM or women with GDM and a family history of diabetes should value antepartum consulting and postpartum followup programmers more than other women with GDM only. Blacks and non-Hispanic Whites could receive more attention, and healthcare providers, especially those in Europe and South-East Asia, could pay more attention to preventive measures. Overall, it is concluded that the RR of diagnosing T2DM after GDM is not directly proportional to GDM prevalence among racial/ethnic groups or geographic regions. Whether the difference is due to lifestyle, genetics or environment needs to be investigated further.