Factors Associated with Gestational Diabetes Mellitus: A Meta-Analysis

Gestational diabetes mellitus (GDM) is a major public health issue, and the aim of the present study was to identify the factors associated with GDM. Databases were searched for observational studies until August 20, 2020. Pooled odds ratios (ORs) were calculated using fixed- or random-effects models. 103 studies involving 1,826,454 pregnant women were identified. Results indicated that maternal age ≥ 25 years (OR: 2.466, 95% CI: (2.121, 2.866)), prepregnancy overweight or obese (OR: 2.637, 95% CI: (1.561, 4.453)), family history of diabetes (FHD) (OR: 2.326, 95% CI: (1.904, 2.843)), history of GDM (OR: 21.137, 95% CI: (8.785, 50.858)), macrosomia (OR: 2.539, 95% CI: (1.612, 4.000)), stillbirth (OR: 2.341, 95% CI: (1.435, 3.819)), premature delivery (OR: 3.013, 95% CI: (1.569, 5.787)), and pregestational smoking (OR: 2.322, 95% CI: (1.359, 3.967)) increased the risk of GDM with all P < 0.05, whereas history of congenital anomaly and abortion, and HIV status showed no correlation with GDM (P > 0.05). Being primigravida (OR: 0.752, 95% CI: (0.698, 0.810), P < 0.001) reduced the risk of GDM. The factors influencing GDM included maternal age ≥ 25, prepregnancy overweight or obese, FHD, history of GDM, macrosomia, stillbirth, premature delivery, pregestational smoking, and primigravida.


Introduction
Gestational diabetes mellitus (GDM), defined as glucose intolerance of variable degree with onset or first recognition during pregnancy, is reported as one of the most common clinical complications of pregnancy [1,2]. According to International Diabetes Federation (IDF) 2017, the prevalence of GDM is expected to be on the rise year by year [3]. Women with GDM may incur a potential risk of adverse outcomes [4,5]. Mothers who have GDM are at risk of developing gestational hypertension and preeclampsia, at risk of suffering from caesarean section, and at risk of inducing subsequent type 2 diabetes mellitus (T2DM) and cardiovascular diseases [6][7][8][9][10][11]. Infants born from GDM women could be prone to abnormal fetal development such as being in macrosomia, having more congenital abnormalities, and having neonatal hypoglycemia [6,12,13]. Consequently, it is suggested that healthcare policy makers should be aware of the significance of GDM for early detection and further intervention.
To date, various relevant factors have been identified as predictors of GDM. Several studies have demonstrated that the frequently reported risk factors of GDM include older maternal ages, prepregnancy obesity, family history of diabetes (FHD) [14,15], previous obstetric outcomes (e.g., macrosomia [16], stillbirth [17], abortion [18], premature delivery [19], congenital anomaly [16], being primigravida [20]), history of GDM [21], infection factors (e.g., Human Immunodeficiency Virus (HIV) [22]), pregestational smoking [23], and socioeconomic factors (educational level, occupation, and monthly household income) [24]. However, there are other evidences suggesting that maternal age, FHD, prepregnancy overweight or obesity, previous history of abortion, stillbirth, and macrosomia showed no significant association with GDM [25,26]. Since most of the information regarding the main factors involved in GDM lack comprehensive analysis, it is necessary to conduct a metaanalysis to further explore the potential factors responsible for GDM.

Materials and Methods
Our study has been approved by the Open Science Framework (OSF) registries (https://osf.io/registries), and the registration number is 10.17605/OSF.IO/4HJGN. This meta-analysis was performed according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) statement. Since this study was based on a meta-analysis of published studies, it did not require patient consent and ethical approval.
2.1. Literature Search Strategy. Four online databases (Web of Science, Embase, PubMed, and Cochrane Library) were systematically searched for articles published till August 20, 2020. We searched PubMed using the following terms: "diabetes mellitus" OR "diabetes" AND "pregnancy" OR "pregnancies" OR "gestation" OR "diabetes, gestational" OR "diabetes, pregnancy-induced" OR "diabetes, pregnancy induced" OR "pregnancy induced diabetes" OR "gestational diabetes" OR "diabetes mellitus gestational" OR "gestational diabetes mellitus" AND "risk factor" OR "risk factors."

Inclusion and Exclusion
Criteria. Inclusion criteria include the following: (1) women with GDM (the observation group) and with healthy pregnancies (the control group); (2) the reported relevant factors in our studies including maternal age ≥ 25 years, prepregnancy overweight or obese, history of GDM, primigravida, history of congenital anomaly, FHD, history of macrosomia, HIV status, history of stillbirth, history of premature delivery, history of abortion, and pregestational smoking; and (3) observational studies.

Data Extraction and Quality
Assessment. The data were extracted by two reviewers (Yu Zhang and Cheng-Ming Xiao) independently according to the inclusion and exclusion criteria. If a conflict existed, the third reviewer (Yi-Meng Gao) would join in extracting the data. The following study features were extracted from each article: the first author's name, year of publication, country, study design, maternal age (years), sample size, the number of GDM cases, and quality assessment scores. The revised Joanna Briggs Institute (JBI) scale was used for cross-sectional studies to evaluate the quality of the literature, with 1-13 being low-risk of bias, and 14-20 being high-risk of bias. The modified Newcastle-Ottawa Scale (NOS) was used for case-control studies and cohort studies, and the studies with scores of 1-4 were considered low quality, while those with scores of 5-10 were considered high quality.

Statistical
Analysis. Data were analyzed using Stata 15.1 software (Stata Corporation, College Station, TX, USA). The factors were assessed by odds ratios (ORs) and 95% confidence intervals (CIs). Heterogeneity tests were performed for each effect size, and random-effects models were adopted when I 2 ≥ 50%; otherwise, fixed effects models were performed. The publication bias was estimated using Egger's test and adjusted by trim and fill method. A difference was considered statistically significant at P < 0:05.

Results
3.1. Literature Search. In this study, 3,586 articles were extracted from PubMed, 5,204 from Embase, 9,340 from Web of Science, 16 from the Cochrane Central, and 7 from other sources. After the removal of duplicate records (n = 13,073), 278 articles were excluded after screening of the titles and abstracts and another 103 through full-text screening for eligibility. Finally, a total of 103 studies (Supplementary Material 1) were included in our study for evaluating the relationship between these factors and GDM. The flow diagram is shown in Figure 1.

Study Characteristics.
A total of 1,826,454 pregnant women were enrolled in this meta-analysis, divided into the observation group (with GDM) composed of 120,696 subjects and the control group (without GDM) composed of 1,705,758 subjects. In terms of the quality of our included studies, scores from the assessment by the revised NOS and JBI scales were summarized in Table 1. The quality scores ranged from 4 to 16. Of the 103 included studies, 29 articles were low quality, while 74 were high quality ( Table 1).

Discussion
In this meta-analysis of 1,826,454 pregnant women from diverse international cohorts, our findings suggested that fac-tors such as maternal age ≥ 25 years, prepregnancy overweight or obese, pregestational smoking, FHD, previous history of GDM, macrosomia, stillbirth, and premature delivery significantly increased the risk of GDM. Besides, being primigravida was associated with a lower risk of GDM, whereas history of congenital anomaly, HIV status, and history of abortion showed no impact on the risk of GDM; controlling these relevant factors for GDM could reduce the serious increase of the occurrence of GDM. Maternal age was reported to be closely associated with GDM. Older maternal age increased the risk of developing GDM, and the threshold for lower risks was recommended as 25 years old by the American Diabetic Association [27], similar to the result of our meta-analysis. However, other studies differed with the result mentioned above, i.e., they recommended that maternal age greater than 35 years was more prone to GDM [20,28]. Although it is shown that there is a certain difference in the cutoff value of maternal age, there is an inevitable risk of developing GDM with the annual increase of age in modern society [29]. The reason for increasing older ages at pregnancy may be related to the implementation of the universal two-child policy, especially in China, as well as a longer period of education and better access to birth control technologies.   Prepregnancy overweight or obese was another major risk factor identified in the current study. A study conducted by Mohan and Chandrakumar also demonstrated that prepregnancy weight management could reduce a woman's risk of GDM [30]. There were other studies with similar results to ours [31,32], despite their varieties of dietary habits and with most people consuming large amounts of alcoholic beverages. Counselling for pregnant women should emphasize the need for women to avoid sedentary lifestyles before pregnancy and to be aware of the risks of GDM to both themselves and the unborn child.
Our study also suggested that FHD (particularly in a firstdegree relative) was strongly related to an increased risk of GDM, which had been observed in a previous study [33]. This was partly because of an increased susceptibility to GDM due to a genetic deficiency in insulin secretion from their first-degree relatives [34]. Therefore, it is important to emphasize that healthcare education providers must obtain accurate personal or family history from their recipients in order to identify at-risk mothers for preventing GDM.
Another significant medical factor associated with a higher risk for developing GDM was history of GDM. Interestingly, a retrospective study [35] and two case-control studies [21,36] also had similar results showing that history of GDM was thought to be a common risk factor in repeated pregnancies [34].
Among the obstetric factors of GDM, Anzaku and Musa pointed out that women with previous history of macrosomia were the only independent risk factor for GDM in the next pregnancy [16], which was similar to our results. A case-control study indicated that women having a history of abortion increased the risk of developing GDM at the   Nhidza (2018) Mohan (2016) Giles (2020) Anzaku (2013) Kong (2020)
With the exception of the most common risk factors, such as maternal age, prepregnancy overweight or obese, FHD, obstetric factors, and infection factors, this study demonstrated that there was a significant correlation between pregestational smoking and GDM. Previous studies have also noted that pregestational smoking was considered to be a risk factor, although its association has been rarely investigated at present [40]. This condition might be explained by the fact that there were several limitations in the way data collection related to smoking was conducted in our study. A recent systematic review examined the relationship between pregestational smoking and the risk of GDM, but no correlation was found [36]. The aspect of smoking in the development of GDM deserves further investigations. Possible uncontrolled confounding factors should be considered, such as the differences in socioeconomic status between groups, selection bias, or even passive smoking.

Conclusions
In our study, maternal age ≥ 25 years, prepregnancy overweight or obese, FHD, previous history of GDM, macrosomia, stillbirth and premature delivery, pregestational smoking, and being primigravida were considered as all independent risk factors of GDM. It is strongly recommended that all pregnant women in the future be screened early for GDM, especially those identified at higher risks of GDM, thereby leading to early diagnosis of GDM and early intervention.

Data Availability
The data used to support the findings of this study are included within the article.

Ethical Approval
Since this study was based on a meta-analysis of published studies, it did not require patient consent and ethical approval.