Early pregnancy prediction of gestational diabetes mellitus risk using prenatal screening biomarkers in nulliparous women

A prediction model for gestational diabetes mellitus based on steroid hormonal changes in early and mid-down syndrome screening: A multicenter longitudinal study

BACKGROUND

Steroid hormones (SH) during pregnancy are associated with the development of gestational diabetes mellitus (GDM). Early and mid-Down syndrome screening is used to assess the risk of Down syndrome in the fetus. It is unclear whether changes in SH during this period can be used as an early predictor of GDM.

METHODS

This study was a multicenter, longitudinal cohort study. GDM is diagnosed by an oral glucose tolerance test (OGTT) between 24 and 28 weeks of gestation. We measured SH levels at early and mid-Down syndrome screening, respectively. Based on the SH changes, logistic regression analysis was used to construct a prediction model for GDM. Finally, evaluated the model's predictive performance by creating a receiver operating characteristic curve (ROC) and performing external validation.

RESULTS

This study enrolled 193 pregnant women (discovery cohort, n = 157; validation cohort, n = 36). SH changes occur dynamically after pregnancy. At early Down syndrome screening, only cortisol (F) (p < 0.05, 95 % CI 4780.95-46083.68) was elevated in GDM. At mid-Down syndrome screening, free testosterone (FT) (p < 0.01, 95 % CI 0.10-0.55) and estradiol (E2) (p < 0.05, 95 % CI 203.55-1784.78) were also significantly elevated. There were significant differences in the rates of change in E2 (Fold change (FC) = 1.3425, p = 0.0072), albumin (ALB) (FC=1.5759, p = 0.0117), and dihydrotestosterone (DHT) (FC=-2.1234, p = 0.0165) between GDM and no-GDM. Stepwise logistic regression analysis resulted in the best predictive model, including six variables (Δweight, ΔF, Δcortisone (E), ΔE2, Δprogesterone (P), ΔDHT). The area under the curve for this model was 0.791, and for the external validation cohort, it was 0.799.

CONCLUSIONS

A GDM prediction model can be constructed using SH measures during early and mid-Down syndrome screening.

Early prediction of gestational diabetes mellitus using first trimester maternal serum pregnancy-associated plasma protein-a: A cross-sectional study

Background and Aims

The oral glucose tolerance test with 75 g glucose is commonly regarded as the gold standard (GS) for the detection of gestational diabetes mellitus (GDM). However, one limitation of this test is its administration in the late second trimester of pregnancy in some countries (e.g., Iran). This study aimed to evaluate the accuracy of pregnancy-associated plasma protein-A (PAPP-A) for predicting GDM in the early first trimester of pregnancy using a novel statistical modeling technique.

Methods

The study population consisted of 344 pregnant women who participated in the first trimester screening program for GDM. Maternal serum PAPP-A levels were measured between 11 and 13 gestational weeks for all participants. A Bayesian latent profile model (LPM) under the skew-t (ST) distribution was employed to estimate the diagnostic accuracy measures of PAPP-A in the absence of GS test outcomes.

Results

The mean (standard deviation) age of the participants was 28.87 ± 5.20 years. The median (interquartile range (IQR)) PAPP-A MoM was 0.91 (0.69-1.34). Utilizing the LPM under the ST distribution while adjusting for covariates, the sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) of PAPP-A were 92% (95% credible interval [CrI]: 0.89, 0.98), 81% (95% CrI: 0.76, 0.91), and 0.91 (95% CrI: 0.83, 0.97), respectively. Notably, the pregnant women with GDM had significantly lower PAPP-A values (β = -0.52, 95% CrI [-0.61, -0.46]).

Conclusion

Generally, our findings confirmed that PAPP-A could serve as a potential screening tool for the identification of GDM in the early stages of pregnancy.

First trimester maternal serum PAPP-A and free β-hCG levels and risk of SGA or LGA in women with and without GDM

BACKGROUND

Maternal gestational diabetes (GDM), small (SGA) and large (LGA) for gestational age neonates are associated with increased morbidity in both mother and child. We studied how different levels of first trimester pregnancy associated plasma protein-A (PAPP-A) and free beta human chorionic gonadotropin (fβ-hCG) were associated with SGA and LGA in GDM pregnancies and controls.

METHODS

Altogether 23 482 women with singleton pregnancies participated in first trimester combined screening and delivered between 2014 and 2018 in Northern Finland and were included in this retrospective case-control study. Women with GDM (n = 4697) and controls without GDM (n = 18 492) were divided into groups below 5th and 10th or above 90th and 95th percentile (pc) PAPP-A and fβ-hCG MoM levels. SGA was defined as a birthweight more than two standard deviations (SD) below and LGA more than two SDs above the sex-specific and gestational age-specific reference mean. Odds ratios were adjusted (aOR) for maternal age, BMI, ethnicity, IVF/ICSI, parity and smoking.

RESULTS

In pregnancies with GDM the proportion of SGA was 2.6% and LGA 4.5%, compared to 3.3% (p = 0.011) and 1.8% (p < 0.001) in the control group, respectively. In ≤ 5th and ≤ 10th pc PAPP-A groups, aORs for SGA were 2.7 (95% CI 1.5-4.7) and 2.2 (95% CI 1.4-3.5) in the GDM group and 3.8 (95% CI 3.0-4.9) and 2.8 (95% CI 2.3-3.5) in the reference group, respectively. When considering LGA, there was no difference in aORs in any high PAPP-A groups. In the low ≤ 5 percentile fβ-hCG MoM group, aORs for SGA was 2.3 (95% CI 1.8-3.1) in the control group. In fβ-hCG groups with GDM there was no association with SGA and the only significant difference was ≥ 90 percentile group, aOR 1.6 (95% CI 1.1-2.5) for LGA.

CONCLUSION

Association with low PAPP-A and SGA seems to be present despite GDM status. High PAPP-A levels are not associated with increased LGA risk in women with or without GDM. Low fβ-hCG levels are associated with SGA only in non-GDM pregnancies.

Identifying Predictor Variables for a Composite Risk Prediction Tool for Gestational Diabetes and Hypertensive Disorders of Pregnancy: A Modified Delphi Study

A composite cardiometabolic risk prediction tool will support the systematic identification of women at increased cardiometabolic risk during pregnancy to enable early screening and intervention. This study aims to identify and select predictor variables for a composite risk prediction tool for cardiometabolic risk (gestational diabetes mellitus and/or hypertensive disorders of pregnancy) for use in the first trimester. A two-round modified online Delphi study was undertaken. A prior systematic literature review generated fifteen potential predictor variables for inclusion in the tool. Multidisciplinary experts (n = 31) rated the clinical importance of variables in an online survey and nominated additional variables for consideration (Round One). An online meeting (n = 14) was held to deliberate the importance, feasibility and acceptability of collecting variables in early pregnancy. Consensus was reached in a second online survey (Round Two). Overall, 24 variables were considered; 9 were eliminated, and 15 were selected for inclusion in the tool. The final 15 predictor variables related to maternal demographics (age, ethnicity/race), pre-pregnancy history (body mass index, height, history of chronic kidney disease/polycystic ovarian syndrome, family history of diabetes, pre-existing diabetes/hypertension), obstetric history (parity, history of macrosomia/pre-eclampsia/gestational diabetes mellitus), biochemical measures (blood glucose levels), hemodynamic measures (systolic blood pressure). Variables will inform the development of a cardiometabolic risk prediction tool in subsequent research. Evidence-based, clinically relevant and routinely collected variables were selected for a composite cardiometabolic risk prediction tool for early pregnancy.

A stepwise prediction and interpretation of gestational diabetes mellitus: Foster the practical application of machine learning in clinical decision

Background

Machine learning has shown to be an effective method for early prediction and intervention of Gestational diabetes mellitus (GDM), which greatly decreases GDM incidence, reduces maternal and infant complications and improves the prognosis. However, there is still much room for improvement in data quality, feature dimension, and accuracy. The contributions and mechanism explanations of clinical data at different pregnancy stages to the prediction accuracy are still lacking. More importantly, current models still face notable obstacles in practical applications due to the complex and diverse input features and difficulties in redeployment. As a result, a simple, practical but accurate enough model is urgently needed.

Design and methods

In this study, 2309 samples from two public hospitals in Shenzhen, China were collected for analysis. Different algorithms were systematically compared to build a robust and stepwise prediction system (level A to C) based on advanced machine learning, and models under different levels were interpreted.

Results

XGBoost reported the best performance with ACC of 0.922, 0.859 and 0.850, AUC of 0.974, 0.924 and 0.913 for the selected level A to C models in the test set, respectively. Tree-based feature importance and SHAP method successfully identified the commonly recognized risk factors, while indicated new inconsistent impact trends for GDM in different stages of pregnancy.

Conclusion

A stepwise prediction system was successfully established. A practical tool that enables a quick prediction of GDM was released at https://github.com/ifyoungnet/MedGDM.This study is expected to provide a more detailed profiling of GDM risk and lay the foundation for the application of the model in practice.

A Prediction Model of Gestational Diabetes Mellitus Based on OGTT in Early Pregnancy: A Prospective Cohort Study

CONTEXT

Gestational diabetes mellitus (GDM) is a common obstetric complication. Although early intervention could prevent the development of GDM, there was no consensus on early identification for women at high risk of GDM.

OBJECTIVE

To develop a reliable prediction model of GDM in early pregnancy.

METHODS

In this prospective cohort study, between May 30, 2021, and August 13, 2022, a total of 721 women were included from Women's Hospital, Zhejiang University School of Medicine. Participants were asked to complete an oral glucose tolerance test (OGTT) during gestational weeks 7 through 14 for early prediction of GDM, and at weeks 24 through 28 for GDM diagnosis. Using OGTT results and baseline characteristics, logistic regression analysis was used to construct the prediction model. Receiver operating characteristic (ROC) curve, Hosmer-Lemeshow test, decision clinical analysis, and a nomogram were used for model performances assessment and visualization. Internal and external validation was performed to testify the stability of this model.

RESULTS

According to the International Association of Diabetes and Pregnancy Study Groups criteria in early OGTT, the mean (SD) age was 30.5 ± 3.7 years in low-risk participants and 31.0 ± 3.9 years in high-risk participants. The area under ROC curve (AUC) of the existing criteria at weeks 7 through 14 varied from 0.705 to 0.724. Based on maternal age, prepregnancy body mass index, and results of early OGTT, the AUC of our prediction model was 0.8720, which was validated by both internal (AUC 0.8541) and external (AUC 0.8241) confirmation.

CONCLUSIONS

The existing diagnostic criteria were unsatisfactory for early prediction of GDM. By combining early OGTT, we provided an effective prediction model of GDM in the first trimester.

Proteins and metabolites fingerprints of gestational diabetes mellitus forming protein-metabolite interactomes are its potential biomarkers

Gestational diabetes mellitus (GDM) is a consequence of glucose intolerance with an inadequate production of insulin that happens during pregnancy and leads to adverse health consequences for both mother and fetus. GDM patients are at higher risk for preeclampsia, and developing diabetes mellitus type 2 in later life, while the child born to GDM mothers are more prone to macrosomia, and hypoglycemia. The universally accepted diagnostic criteria for GDM are lacking, therefore there is a need for a diagnosis of GDM that can identify GDM at its early stage (first trimester). We have reviewed the literature on proteins and metabolites fingerprints of GDM. Further, we have performed protein-protein, metabolite-metabolite, and protein-metabolite interaction network studies on GDM proteins and metabolites fingerprints. Notably, some proteins and metabolites fingerprints are forming strong interaction networks at high confidence scores. Therefore, we have suggested that those proteins and metabolites that are forming protein-metabolite interactomes are the potential biomarkers of GDM. The protein-metabolite biomarkers interactome may help in a deep understanding of the prognosis, pathogenesis of GDM, and also detection of GDM. The protein-metabolites interactome may be further applied in planning future therapeutic strategies to promote long-term health benefits in GDM mothers and their children.

Study on the Relationship and Predictive Value of First-Trimester Pregnancy-Associated Plasma Protein-A, Maternal Factors, and Biochemical Parameters in Gestational Diabetes Mellitus: A Large Case-Control Study in Southern China Mothers

OBJECTIVE

To investigate the relationship and predictive value of first-trimester pregnancy-associated plasma protein A (PAPP-A), maternal factors, and biochemical parameters with gestational diabetes mellitus (GDM) in southern China mothers.

METHODS

This study recruited 4872 pregnant women. PAPP-A, the free beta subunit of human chorionic gonadotropin (free β-HCG), fasting plasma glucose (FPG), total cholesterol (TC), triglycerides (TG), and high- and low-density lipoproteins (HDL, LDL) were measured at 11-13⁺ weeks of gestation. GDM was diagnosed based on a 75 g oral glucose tolerance test at 24-28 weeks of gestation. We performed stepwise logistic regression analysis to determine the odds ratio (OR) and the 95% confidence interval (CI) of GDM. We used Receiver Operating Characteristic (ROC) curves with the area under the curve (AUC) to evaluate the predictive value of PAPP-A, maternal factors, and biochemical markers. The significance of the differences between the AUC values was assessed using the DeLong test.

RESULTS

GDM was diagnosed in 750 (15.39%) women. Independent factors for GDM were age, pre-gestational BMI, GWG before a diagnosis of GDM, previous history of GDM, family history of diabetes, FPG, TG, LDL, PAPP-A, and TC. The AUC of PAPP-A was 0.56 (95% CI 0.53-0.58). The AUC of a model based on combined maternal factors, biochemical markers, and PAPP-A was 0.70 (95% CI 0.68-0.72). Differences in AUC values between PAPP-A alone and the model based on combined maternal factors, biochemical markers, and PAPP-A were statistically significant (Z= 9.983, P<0.001).

CONCLUSION

A Low serum PAPP-A level in the first trimester is an independent risk factor for developing GDM later in pregnancy. However, it is not a good independent predictor although the predictive value of a low serum PAPP-A level increases when combined with maternal factors and biochemical markers.

Cardiovascular Disease-Associated MicroRNAs as Novel Biomarkers of First-Trimester Screening for Gestational Diabetes Mellitus in the Absence of Other Pregnancy-Related Complications

We assessed the diagnostic potential of cardiovascular disease-associated microRNAs for the early prediction of gestational diabetes mellitus (GDM) in singleton pregnancies of Caucasian descent in the absence of other pregnancy-related complications. Whole peripheral venous blood samples were collected within 10 to 13 weeks of gestation. This retrospective study involved all pregnancies diagnosed with only GDM (n = 121) and 80 normal term pregnancies selected with regard to equality of sample storage time. Gene expression of 29 microRNAs was assessed using real-time RT-PCR. Upregulation of 11 microRNAs (miR-1-3p, miR-20a-5p, miR-20b-5p, miR-23a-3p, miR-100-5p, miR-125b-5p, miR-126-3p, miR-181a-5p, miR-195-5p, miR-499a-5p, and miR-574-3p) was observed in pregnancies destinated to develop GDM. Combined screening of all 11 dysregulated microRNAs showed the highest accuracy for the early identification of pregnancies destinated to develop GDM. This screening identified 47.93% of GDM pregnancies at a 10.0% false positive rate (FPR). The predictive model for GDM based on aberrant microRNA expression profile was further improved via the implementation of clinical characteristics (maternal age and BMI at early stages of gestation and an infertility treatment by assisted reproductive technology). Following this, 69.17% of GDM pregnancies were identified at a 10.0% FPR. The effective prediction model specifically for severe GDM requiring administration of therapy involved using a combination of these three clinical characteristics and three microRNA biomarkers (miR-20a-5p, miR-20b-5p, and miR-195-5p). This model identified 78.95% of cases at a 10.0% FPR. The effective prediction model for GDM managed by diet only required the involvement of these three clinical characteristics and eight microRNA biomarkers (miR-1-3p, miR-20a-5p, miR-20b-5p, miR-100-5p, miR-125b-5p, miR-195-5p, miR-499a-5p, and miR-574-3p). With this, the model identified 50.50% of GDM pregnancies managed by diet only at a 10.0% FPR. When other clinical variables such as history of miscarriage, the presence of trombophilic gene mutations, positive first-trimester screening for preeclampsia and/or fetal growth restriction by the Fetal Medicine Foundation algorithm, and family history of diabetes mellitus in first-degree relatives were included in the GDM prediction model, the predictive power was further increased at a 10.0% FPR (72.50% GDM in total, 89.47% GDM requiring therapy, and 56.44% GDM managed by diet only). Cardiovascular disease-associated microRNAs represent promising early biomarkers to be implemented into routine first-trimester screening programs with a very good predictive potential for GDM.

Optimising Cardiometabolic Risk Factors in Pregnancy: A Review of Risk Prediction Models Targeting Gestational Diabetes and Hypertensive Disorders

Cardiovascular disease, especially coronary heart disease and cerebrovascular disease, is a leading cause of mortality and morbidity in women globally. The development of cardiometabolic conditions in pregnancy, such as gestational diabetes mellitus and hypertensive disorders of pregnancy, portend an increased risk of future cardiovascular disease in women. Pregnancy therefore represents a unique opportunity to detect and manage risk factors, prior to the development of cardiovascular sequelae. Risk prediction models for gestational diabetes mellitus and hypertensive disorders of pregnancy can help identify at-risk women in early pregnancy, allowing timely intervention to mitigate both short- and long-term adverse outcomes. In this narrative review, we outline the shared pathophysiological pathways for gestational diabetes mellitus and hypertensive disorders of pregnancy, summarise contemporary risk prediction models and candidate predictors for these conditions, and discuss the utility of these models in clinical application.

Second trimester prediction of gestational diabetes: maternal analytes as an additional screening tool

OBJECTIVES

Early diagnosis of gestational diabetes can lead to greater optimization of glucose control. We evaluated associations between maternal serum analytes (alpha-fetoprotein [AFP], free beta-human chorionic gonadotropin [beta-hCG], inhibin, and estriol) and the development of gestational diabetes mellitus (GDM).

METHODS

This retrospective cohort study identified single-ton pregnancies with available second trimester serum analytes between 2009 and 2017. GDM was identified by ICD-9 and -10 codes. We examined the associations between analyte levels and GDM and to adjust for potential confounders routinely collected during genetic serum screening (maternal age, BMI, and race) using logistic regression. Optimal logistic regression predictive modeling for GDM was then performed using the analyte levels and the above mentioned potential confounders. The performance of the model was assessed by receiver operator curves.

RESULTS

Out of 5,709 patients, 660 (11.6%) were diagnosed with GDM. Increasing AFP and estriol were associated with decreasing risk of GDM, aOR 0.76 [95% CI 0.60-0.95] and aOR 0.67 [95% CI 0.50-0.89] respectively. Increasing beta-hCG was associated with a decreasing risk for GDM(aOR 0.84 [95% CI 0.73-0.97]). There was no association with inhibin. The most predictive GDM predictive model included beta-hCG and estriol in addition to the clinical variables of age, BMI, and race (area under the curve (AUC 0.75), buy this was not statistically different than using clinical variables alone (AUC 0.74) (p=0.26).

CONCLUSIONS

Increasing second trimester AFP, beta-hCG, and estriol are associated with decreasing risks of GDM, though do not improve the predictive ability for GDM when added to clinical risk factors of age, BMI, and race.

Steroid profiling and genetic variants in Chinese women with gestational diabetes mellitus

Previous studies have demonstrated that steroids were associated with gestational diabetes mellitus (GDM). However, results from different studies remained inconsistent, and only a limited range of steroids were investigated in these studies. Therefore, we aimed to analyze comprehensive steroid profiling in Chinese women with GDM during third-trimester pregnancy. In 97 Chinese pregnant women, we measured steroid profile using a LC-MS/MS method, and calculated product-to-precursor ratios in metabolic pathways of steroids. Then sixteen genetic variants of genes encoding steroidogenic enzymes were genotyped by MassARRAY system. There were significant differences (P < 0.05) and obvious changes (fold change <0.67 or>1.5) in steroids (testosterone, estriol, pregnenolone and dehydroepiandrosterone) and product-to-precursor ratios (E2/T and T/AD) between GDM and control groups. After adjusting for maternal age, the TT genotype and T allele of CYP19A1 rs10046 were associated with an increased risk of GDM. And the CC genotype and C allele of HSD17B3 rs2257157 were also associated with an increased risk of GDM. Besides, pregnant women carrying TT genotype of CYP19A1 rs10046 and CC genotype of HSD17B3 rs2257157 had a lower E2/T ratio and higher T/AD ratio respectively comparing with those carrying other genotypes. In conclusion, our study suggested that testosterone, estriol, pregnenolone and dehydroepiandrosterone might be differential metabolites for gestational diabetes mellitus. The genetic variants rs10046 of CYP19A1 and rs2257157 of HSD17B3 could predispose to GDM in Chinese women.

Emerging Protein Biomarkers for the Diagnosis or Prediction of Gestational Diabetes-A Scoping Review

Introduction: Gestational diabetes (GDM), defined as hyperglycemia with onset or initial recognition during pregnancy, has a rising prevalence paralleling the rise in type 2 diabetes (T2DM) and obesity. GDM is associated with short-term and long-term consequences for both mother and child. Therefore, it is crucial we efficiently identify all cases and initiate early treatment, reducing fetal exposure to hyperglycemia and reducing GDM-related adverse pregnancy outcomes. For this reason, GDM screening is recommended as part of routine pregnancy care. The current screening method, the oral glucose tolerance test (OGTT), is a lengthy, cumbersome and inconvenient test with poor reproducibility. Newer biomarkers that do not necessitate a fasting sample are needed for the prompt diagnosis of GDM. The aim of this scoping review is to highlight and describe emerging protein biomarkers that fulfill these requirements for the diagnosis of GDM. Materials and Methods: This scoping review was conducted according to preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines for scoping reviews using Cochrane Central Register of Controlled Trials (CENTRAL), the Cumulative Index to Nursing & Allied Health Literature (CINAHL), PubMed, Embase and Web of Science with a double screening and extraction process. The search included all articles published in the literature to July 2020. Results: Of the 3519 original database citations identified, 385 were eligible for full-text review. Of these, 332 (86.2%) were included in the scoping review providing a total of 589 biomarkers studied in relation to GDM diagnosis. Given the high number of biomarkers identified, three post hoc criteria were introduced to reduce the items set for discussion: we chose only protein biomarkers with at least five citations in the articles identified by our search and published in the years 2017-2020. When applied, these criteria identified a total of 15 biomarkers, which went forward for review and discussion. Conclusions: This review details protein biomarkers that have been studied to find a suitable test for GDM diagnosis with the potential to replace the OGTT used in current GDM screening protocols. Ongoing research efforts will continue to identify more accurate and practical biomarkers to take GDM screening and diagnosis into the 21st century.

Metabolomic profile associated with obstructive sleep apnoea severity in obese pregnant women with gestational diabetes mellitus: A pilot study

Obstructive sleep apnoea (OSA) is prevalent in obese women with gestational diabetes mellitus (GDM). The present pilot study explored associations between OSA severity and metabolites in women with GDM. A total of 81 obese women with diet-controlled GDM had OSA assessment (median gestational age [GA] 29 weeks). The metabolic profile was assayed from fasting serum samples via liquid chromatography-mass spectrometry (LC-MS) using an untargeted approach. Metabolites were extracted and subjected to an Agilent 1,290 UPLC coupled to an Agilent 6,545 quadrupole time-of-flight (Q-TOF) MS. Data were acquired using electrospray ionisation in positive and negative ion modes. The raw LC-MS data were processed using the OpenMS toolkit to detect and quantify features, and these features were annotated using the Human Metabolite Database. The feature data were compared with OSA status, apnea-hypopnea index (AHI), body mass index (BMI) and GA using "limma" in R. Correlation analyses of the continuous covariates were performed using Kendall's Tau test. The p values were adjusted for multiple testing using the Benjamini-Hochberg false discovery rate correction. A total of 42 women (51.8%) had OSA, with a median AHI of 9.1 events/hr. There were no significant differences in metabolomics profiles between those with and without OSA. However, differential analyses modelling in GA and BMI found 12 features that significantly associated with the AHI. These features could be annotated to oestradiols, lysophospholipids, and fatty acids, with higher levels related to higher AHI. Metabolites including oestradiols and phospholipids may be involved in pathogenesis of OSA in pregnant women with GDM. A targeted approach may help elucidate our understanding of their role in OSA in this population.

Machine Learning Prediction Models for Gestational Diabetes Mellitus: A meta- analysis (Preprint)

Background

Gestational diabetes mellitus (GDM) is a common endocrine metabolic disease, involving a carbohydrate intolerance of variable severity during pregnancy. The incidence of GDM-related complications and adverse pregnancy outcomes has declined, in part, due to early screening. Machine learning (ML) models are increasingly used to identify risk factors and enable the early prediction of GDM.

Objective

The aim of this study was to perform a meta-analysis and comparison of published prognostic models for predicting the risk of GDM and identify predictors applicable to the models.

Methods

Four reliable electronic databases were searched for studies that developed ML prediction models for GDM in the general population instead of among high-risk groups only. The novel Prediction Model Risk of Bias Assessment Tool (PROBAST) was used to assess the risk of bias of the ML models. The Meta-DiSc software program (version 1.4) was used to perform the meta-analysis and determination of heterogeneity. To limit the influence of heterogeneity, we also performed sensitivity analyses, a meta-regression, and subgroup analysis.

Results

A total of 25 studies that included women older than 18 years without a history of vital disease were analyzed. The pooled area under the receiver operating characteristic curve (AUROC) for ML models predicting GDM was 0.8492; the pooled sensitivity was 0.69 (95% CI 0.68-0.69; P<.001; I²=99.6%) and the pooled specificity was 0.75 (95% CI 0.75-0.75; P<.001; I²=100%). As one of the most commonly employed ML methods, logistic regression achieved an overall pooled AUROC of 0.8151, while non–logistic regression models performed better, with an overall pooled AUROC of 0.8891. Additionally, maternal age, family history of diabetes, BMI, and fasting blood glucose were the four most commonly used features of models established by the various feature selection methods.

Conclusions

Compared to current screening strategies, ML methods are attractive for predicting GDM. To expand their use, the importance of quality assessments and unified diagnostic criteria should be further emphasized.