Gestational diabetes mellitus in women born small or premature: Systematic review and meta-analysis

Searches

We searched databases such as MEDLINE, Embase, and CINAHL and study registries including ClinicalTrials.gov and ICTRP from launch until 29 October 2020. Qualified authors (YT and YK) developed the search strategy (Supplementary Table S2). No language or publication status restrictions were imposed. Reference lists of shortlisted studies were searched manually for additional potentially eligible titles.

Study selection

Studies were eligible for inclusion if they were observational cohort or case-control studies. Case reports or series were excluded from the present review. We included studies that involved pregnant women regardless of study setting. The exposures of interest were the infancy parameters of presently pregnant women and were defined as follows: LBW, birth weight <2500 g ⁷; small for gestational age (SGA), birth weight <10th percentile for the given gestational age, stratified by sex, using the average weight of gestational age ¹⁶; and preterm birth, gestational age of <37 weeks ¹⁷. When data on both birth weight and gestational age were reported, we extracted data on birth weight in preference. The comparator group comprised women who were not born small or born at full term.

The outcome of interest was GDM, as defined by the International Association of Diabetes Pregnancy Study Groups (IADPSG), World Health Organization (WHO), American Diabetes Association or Endocrine Society, or International Classification of Diseases 11th revision (ICD-11) or earlier ^18–22. If studies used other definitions, they were included in the present review; however, we removed them to assess the robustness of the pooled estimates. For studies that reported LBW, preterm birth, or SGA as a risk factor in pregnant women without reporting the association with GDM, we contacted study authors to acquire estimates of such associations, where available. These additional estimates were included in the present analysis, provided they were measures of an association between at least one of the exposure factors and the outcome of interest.

Two investigators independently screened article titles and abstracts to shortlist relevant studies; subsequently, the same sets of authors assessed the full text for study eligibility. In cases where data were incomplete and precluded study eligibility assessment, we contacted study authors with requests for clarification. Multiple publications were assessed together; the record with the most complete data was included in the present review.

Data extraction and quality assessment

Two investigators independently extracted data from all included studies, using a pilot-tested, uniform data extraction sheet. Any discrepancies between reviewers were resolved through consensus between two reviewers or arbitration by a third reviewer, as required. For studies that compared three or more exposure groups, we contacted study authors to obtain data comparing two groups of interest. In cases where this approach was unsuitable, we extracted the relevant data, as reported, and performed subgroup comparisons between the two groups subdivided by specific thresholds (i.e., birth weight 2500 g, <10th percentile, and gestational age 37 weeks for LBW, SGA, and preterm birth, respectively), as this approach may have resulted in conservative effect estimates. The same authors who performed data extraction also independently assessed the risk of bias in each study, using the Quality In Prognosis Studies (QUIPS) tool ²³. We prospectively identified the following candidate confounders: age, obesity, smoking status, socioeconomic status, diabetes mellitus before the index gestation, and family history of diabetes ^{24, 25}.

Data synthesis and analysis

We obtained pooled and adjusted ORs with 95% CI estimates of GDM for the exposure and control groups using the DerSimonian and Laird random-effects method. We calculated the absolute risk difference for GDM between the exposure and control groups in low-(control group: GDM risk was assumed to be 2.0%), moderate- (10%), and high- (20%) prevalence groups, using the pooled odds ratios (OR) and 95% confidence intervals (CIs). This assumption was made based on a previous report and our clinical expertise ²⁶.

Publication bias was assessed qualitatively by visual inspection of the funnel plot and quantitatively by Egger’s test ²⁷. Where asymmetry was observed in the funnel plot, we investigated the likely source of this asymmetry using the contour-enhanced funnel plot. We evaluated between-study heterogeneity visually, using forest plots, and quantitatively, using I² and τ² statistics. We used the Cochrane chi-square test to calculate I² and τ² statistics. We performed a pre-specified subgroup analysis based on types of exposure (preterm birth, LBW, or SGA). In pre-specified sensitivity analyses, we used crude ORs instead of adjusted ORs and excluded studies using non-standard definitions of GDM. Some studies assessed the risk of GDM among women born with a weight >4000 g (macrosomia); these studies were excluded from post-hoc sensitivity analysis, as a previous review has shown a U-shaped association between mother’s birth weight and GDM risk ¹⁰.

All analyses were performed using STATA 14.2 (StataCorp LP, Texas) and RevMan 5.4 (Cochrane Collaboration, UK). Two-sided p-values <0.05 were considered indicative of statistical significance.

Prematurity and size at birth and the risk of gestational diabetes mellitus

The median GDM rate in the control groups of the included cohort studies was 2.9% (range: 0.5% to 22%). Figure 2 presents a forest plot summarising the studies that assessed the association between preterm birth or size at birth with GDM. Premature birth, LBW, and SGA status were associated with a higher GDM risk (pooled OR, 1.84; 95% CI: 1.54 to 2.20; I² = 78.3%; τ² = 0.07). Supplementary Table S5 summarises the absolute risk difference in pregnant women born with LBW, SGA status, or born preterm in the low- (2.0% risk of GDM in the control group), medium- (10%), and high- (20%) GDM prevalence groups. The absolute risk increases were 1.6% (95% CI: 1.0 to 2.1%), 7.0% (95% CI: 4.6 to 9.6%), and 11.5% (95% CI: 7.8 to 15.5%) in low-, moderate-, and high-prevalence settings, respectively. The certainty of evidence was low due to serious concerns of risk of bias and publication bias.

• Download figure
• Open in new tab

Figure 2. Risk of gestational diabetes among women born preterm, with low birth weight, or small-for-gestational-age status

Effect size (ES, represented as adjusted odds ratios); CI, confidence interval. ES was determined using the random-effects model weighted by the inverse of the variance estimate. Squares represent ES, with marker size reflecting the statistical weight of the study, obtained using random-effects meta-analysis; horizontal lines represent 95% CIs; diamonds represent the subgroup and overall odds ratios and 95% CIs for gestational diabetes.

Figure 3 presents study estimates in a funnel plot. The plot appeared asymmetrical, and Egger’s test for funnel plot asymmetry was statistically significant (p-value = 0.030). Supplementary Figure S2 shows the contour-enhanced funnel plot, which suggests the existence of some missing studies on the left-hand side of the plot; these studies would have yielded statistically non-significant findings.

• Download figure
• Open in new tab

Figure 3. Funnel plot for the evaluation of publication bias

The solid vertical line represents the summary estimate of the association between preterm birth, low birth weight, and small for gestational age status and gestational diabetes (using random-effects meta-analysis). A significant publication bias was detected (p□=□0.030 for Egger’s test). The funnel plot shows asymmetry, which indicates publication bias.

Data on the birth weight of mothers with or without GDM, obtained from three studies excluded from the meta-analysis, are presented in Supplementary Table S5. These studies consistently reported that mothers with GDM were born with lower birth weights than those without GDM.

Subgroup and sensitivity analyses

There was substantial between-study heterogeneity (I² = 78.3%). Figure 2 presents the results of subgroup analyses for the types of exposure (LBW, preterm birth, or SGA). Although all types of exposure were associated with GDM, there was significant heterogeneity due to the type of exposure (p for interaction = 0.004). The results of additional sensitivity analyses are presented in Supplementary Figure S3, Figure S4, and Figure S5.

Main Findings

We found that women born small or premature may have future risk of GDM. However, the evidence certainty was low, and the presented findings may be overestimated, as we observed some evidence of publication bias. These findings were approximately consistent across the subgroups, including different populations, exposures, and studies of varied methodological quality; these findings were robust in sensitivity analyses.

Our finding that the mother’s size at birth or premature birth may affect GDM risk was consistent with that of a previous narrative review ¹⁰. The strength of this association was similar to that observed in women with a family history of diabetes mellitus, an established risk factor for GDM ²⁹. However, the importance of the risk factor in clinical decision depends on the absolute risk difference. Our findings suggested that careful review of the mother’s birth status may indicate her risk of GDM and guide pregnancy management in moderate to high prevalence settings. The mother’s preterm birth status and size at birth are not currently considered risk factors for GDM in any of the major guidelines or risk models ^{30, 31}. Our findings may help further refine these guidelines and models or to develop new ones.

The certainty of evidence for the association between premature birth or SGA status and GDM was low due to the high risk of publication bias, as shown by funnel-plot analysis. The contour-enhanced funnel plot suggests that studies with non-significant findings may not have been published. Although we did not identify any ongoing or unpublished studies, this did not eliminate the risk of publication bias, as observational studies are less likely to be registered than clinical trials ³². Thus, the reported estimates may be overestimates. The studies included in this review tended not to adjust for confounders, such as smoking, obesity, socioeconomic status, and family history of diabetes. Future studies should adjust for these factors.

The main result of this review was subject to substantial between-study heterogeneity, as shown by the I² statistic ¹³. This heterogeneity may be due to the different types of exposure (LBW, SGA, or preterm birth) considered in this study. However, as all exposure types were associated with increased GDM risk, the high I² statistic may be due to the large number of participants and narrow CIs of the primary studies ³³. Given these findings, we did not assign a low rating to the inconsistency domain of the GRADE criteria ¹⁵.

The underlying mechanism of the association between preterm birth or SGA status and subsequent GDM may be gestational malnutrition due to maternal malnutrition or placental insufficiency ³⁴. Findings from animal studies have suggested that malnutrition in utero is associated with reduced β-cell counts, pancreas weight, and pancreatic insulin content ³⁵. According to the Barker foetal origin hypothesis, these foetal programming events may affect the future risk of disease ⁹. A review of epidemiological studies has suggested that LBW and preterm birth are associated with the risk of type 2 diabetes in adulthood; a similar mechanism is possible for GDM ³⁶.

Strengths and Limitations

A key strength of this review is that it is the first to provide a comprehensive summary of evidence on the association between birth size or premature birth and future GDM risk. This study followed the methodological recommendations presented in the Cochrane Handbook, MOOSE guidelines, and GRADE criteria ^13–15. Moreover, this study included previously unpublished data and a large sample size.

Nevertheless, this study has some limitations. First, the included studies were old and may not represent the current clinical practice. The definition of GDM proposed by the IADPSG in 2010 has resulted in an increase in GDM prevalence ³⁷. For example, the prevalence of GDM in the United States increased from 4.6% in 2006 to 8.2% in 2016 ³⁸. The median prevalence of GDM in the control groups of the included studies was 2.9%. However, empirical evidence suggests that relative effect measures are, on average, consistent across different settings; in the present study, we estimated absolute risk differences separately for low-, moderate-, and high-prevalence settings ³⁹. Second, 5 of 15 studies divided birth size and preterm birth categories into three or more comparative groups, which could not be combined into two comparison groups of interest. This lack of data required methodological adjustments, as described previously. Lastly, this review only assessed certainty in estimates of association between prognostic factors and an outcome. Future studies are required to determine whether these factors can help risk-stratify pregnant women and improve the clinical management of GDM.