Background. Gestational diabetes mellitus (GDM) contributes to adverse pregnancy and birth outcomes. In recent decades, extensive research has been devoted to the early prediction of GDM by various methods. Machine learning methods are flexible prediction algorithms with potential advantages over conventional regression. Objective. The purpose of this study was to use machine learning methods to predict GDM and compare their performance with that of logistic regressions. Methods. We performed a retrospective, observational study including women who attended their routine first hospital visits during early pregnancy and had Down’s syndrome screening at 16-20 gestational weeks in a tertiary maternity hospital in China from 2013.1.1 to 2017.12.31. A total of 22,242 singleton pregnancies were included, and 3182 (14.31%) women developed GDM. Candidate predictors included maternal demographic characteristics and medical history (maternal factors) and laboratory values at early pregnancy. The models were derived from the first 70% of the data and then validated with the next 30%. Variables were trained in different machine learning models and traditional logistic regression models. Eight common machine learning methods (GDBT, AdaBoost, LGB, Logistic, Vote, XGB, Decision Tree, and Random Forest) and two common regressions (stepwise logistic regression and logistic regression with RCS) were implemented to predict the occurrence of GDM. Models were compared on discrimination and calibration metrics. Results. In the validation dataset, the machine learning and logistic regression models performed moderately (AUC 0.59-0.74). Overall, the GBDT model performed best (AUC 0.74, 95% CI 0.71-0.76) among the machine learning methods, with negligible differences between them. Fasting blood glucose, HbA1c, triglycerides, and BMI strongly contributed to GDM. A cutoff point for the predictive value at 0.3 in the GBDT model had a negative predictive value of 74.1% (95% CI 69.5%-78.2%) and a sensitivity of 90% (95% CI 88.0%-91.7%), and the cutoff point at 0.7 had a positive predictive value of 93.2% (95% CI 88.2%-96.1%) and a specificity of 99% (95% CI 98.2%-99.4%). Conclusion. In this study, we found that several machine learning methods did not outperform logistic regression in predicting GDM. We developed a model with cutoff points for risk stratification of GDM.

This study was designed to examine the impact of prenatal fine particulate matter (PM2.5) exposure on fetal growth and the underlying placental epigenetic mechanism in a cohort of Chinese women. Within the prospective Shanghai Mother-Child Pairs cohort (Shanghai MCPC), 329 women carrying singleton pregnancy with a due date in 2018 were recruited between 2017 and 2018. Maternal PM2.5 exposure levels were estimated using gestational exposure prediction model combining satellite-driven ambient concentrations and personal air sampling. Fetal growth characteristics were evaluated by prenatal ultrasound examinations and anthropometric measurements at birth. In a discovery phase, whole-genome DNA methylation analysis was performed using the Infinium 850 K array. In a validation phase, placental DNA methylation was measured using bisulfite pyrosequencing for five candidate genes that showed the most significant alterations and function relevance in our methylation array screen, including BID (BH3 interacting domain death agonist), FOXN3 (Forkhead box N3), FOXP1 (Forkhead box P1), IGF2 (Insulin-like growth factor 2) and HSD11B2 (Hydroxysteroid 11-beta dehydrogenase 2). Multivariate linear regression models were applied to examine the associations among PM2.5 exposure, fetal growth characteristics and DNA methylation on placental candidate genes. Sobel tests were used to evaluate the mediating role of DNA methylation in multivariable models. After excluding women who withdrew or failed to provide placenta, a total of 287 pregnant women with an average age of 30 entered the final analysis. Increased PM2.5 exposure was significantly associated with reduced biparietal diameter (BPD) (β: -0.136 mm, 95% CI: -0.228 to -0.043), head circumference (HC) (β: -0.462 mm, 95% CI: -0.782 to -0.142), femur length (FL) (β: -0.113 mm, 95% CI: -0.185 to -0.041) and abdominal circumference (AC) (β: -0.371 mm, 95% CI: -0.672 to -0.071) in the second trimester and birth length (β: -0.013 cm, 95% CI: -0.025 to -0.001). Prenatal PM2.5 exposure could lead to aberrant changes in DNA methylation profile of placenta genome, which were mainly enriched in reproductive development, energy metabolism and immune response. DNA methylation of IGF2 and BID showed significant associations with PM2.5 exposures during all exposure windows. In addition, BID methylation was negatively correlated with HC (β: -1.396 mm, 95% CI: -2.582 to -0.209) and BPD (β: -0.330 mm, 95% CI: -0.635 to -0.026) in the second trimester. Further mediation analysis indicated that BID methylation mediated about 30% of the effects of PM2.5 exposure on HC. These findings collectively suggested that prenatal PM2.5 exposure may cause adverse effects on fetal growth by modifying placental DNA methylation.

Polybrominated diphenyl ethers (PBDEs) are environmental chemicals with harmful effects on pregnancy, but their effects on adverse developmental outcomes are not fully understood. The placental DNA methylation is strongly influenced by prenatal environmental factors and has been linked to fetal growth. To evaluate the association between in utero PBDEs exposure, placental DNA methylation changes (growth regulatory genes), and pathophysiology of fetal growth (birth outcomes, fetal growth retardation) in a population-based pregnancy cohort study. This was a nested case-control study within the prospective Wenzhou Birth Cohort including 130 fetal growth retardation (FGR) cases and 130 healthy controls and their mothers recruited from June 2016 to June 2017. FGR was diagnosed based on the comprehensive evaluation of ultrasound results at 24, 28, and 32 weeks of gestation. Neonatal birth measurements were obtained from medical records. Gestational exposure to 19 PBDEs, including 13 lower BDE congeners (BDE-17-190) and 6 higher brominated BDE congeners (BDE-196-209), were determined by gas chromatography tandem mass spectrometry in the umbilical cord blood. Placental DNA methylation changes of one repetitive element (LINE1) and two candidate genes (HSD11B2, IGF2) were characterized by quantitative polymerase chain reaction-pyrosequencing. Multiple linear regression and logistic regression models were used to examine the associations among PBDEs exposure, fetal growth indicators, and DMR (differential methylation region) methylation fractions. Sobel tests were conducted to assess DNA methylation as a mediator in multivariate models. After excluding women who withdrew from the study or were lost to follow-up or failed to provide placenta or umbilical cord blood, 249 mother-newborn pairs (124 FGR cases, 125 controls) were included in the final analysis. Elevated BDE-206 (OR: 1.569, 95% CIs: 1.053-2.338), BDE-17-190 (OR: 2.860, 95% CIs: 1.233-6.634), BDE-196-209 (OR: 1.688, 95% CIs: 1.024-2.783) and ∑19PBDEs (OR: 2.387, 95% CIs: 1.220-4.668) concentrations were associated with increased risk of FGR in newborns. FGR birth was also associated with increased DNA methylation of HSD11B2 (OR: 1.145, 95% CIs: 1.032-1.270) and decreased DNA methylation of IGF2 (OR: 0.892, 95% CIs: 0.845-0.941). In addition, BDE-17-190 showed significant associations with DNA methylation of HSD11B2 and IGF2 (β: 1.127, 95% CIs: 0.069-2.186; β: -3.452, 95% CIs: -5.512-1.392), indicating placental DNA methylation changes of HSD11B2 and IGF2 were related to both lower BDE congeners exposure and fetal growth. Further mediation analyses showed that IGF2 methylation mediated about 40% of the effects of BDE-17-190 in umbilical cord blood on neonatal FGR. We report an inverse association between in utero exposures to PBDEs and fetal growth and provide evidence supporting epigenetic gene plasticity in these associations. Changes in placental DNA methylation might be part of the underlying biological pathway between prenatal PBDEs exposure and adverse fetal growth.

Context Universal early-pregnancy screening for overt diabetes reveals intermediate hyperglycemia (fasting plasma glucose [FPG] [5.1-6.9 mM]). Objective We evaluated the association between early-pregnancy intermediate hyperglycemia and adverse pregnancy outcomes among women without gestational diabetes. Methods This retrospective cohort study was conducted at the Obstetrics and Gynecology Hospital, Shanghai, China, from 2013 to 2017. All singleton pregnancies with FPG less than or equal to 6.9 mM in early pregnancy and receiving a 75-g oral glucose tolerance test (OGTT) were included. Women with prepregnancy diabetes were excluded. Individuals with normal OGTT were analyzed. Pregnancy outcomes for FPG less than 5.1 mM and intermediate hyperglycemia were evaluated. The primary outcomes were large for gestational age (LGA) and primary cesarean delivery. Multivariate logistic regressions were conducted. Statistical significance was defined as P less than .05. Results In total, 24 479 deliveries were included, of which 23 450 (95.8%) had normal OGTTs later in pregnancy (NGT). There were 807 (3.4%) women who had an FPG of 5.1 to 6.9 mM in early pregnancy. Compared to the NGT group with an FPG of less than 5.1 mM in early pregnancy (N = 20692), the intermediate hyperglycemia NGT group (N = 693) had a higher age and body mass index (BMI), and significantly higher rates of LGA, primary cesarean delivery, preterm birth, preeclampsia, and neonatal distress. The rates of primary cesarean delivery (adjusted odds ratio [AOR] 1.24; 95% CI, 1.05-1.45), preterm birth (AOR 1.75; 95% CI, 1.29-2.36), and neonatal distress (AOR 3.29; 95% CI, 1.57-6.89) remained statistically significantly higher after adjustments for maternal age, BMI, and other potential confounding factors. Conclusion Women with intermediate hyperglycemia in early pregnancy are at an increased risk for adverse maternal-fetal outcomes, even with normal future OGTTs.