Statistical analyses
Continuous data are presented as mean ± standard deviation (SD) for normally distributed variables and as medians and interquartile ranges (IQRs) for non-normally distributed variables. Categorical data are presented as frequencies and percentages. Student’s t-test, Mann–Whitney U test, chi-square test, and Fisher’s exact test were used to compare baseline characteristics between the normal and GDM groups. Latent profile analysis (LPA) was performed to identify glucose patterns in patients with GDM based on four measurements during the OGTT. This method assumes that unobserved latent profiles generate patterns of responses in a series of continuous variables. The optimal number of clusters was determined by considering the Bayesian information criterion (BIC) value, distribution of cluster membership probabilities, cluster sizes, and interpretability of the identified patterns.19,20 A three-cluster model was selected because it had a lower BIC value than the other models, and all cluster sizes were >10% of the number of patients with GDM. To classify individuals exclusively into three glucose patterns, we assigned patients to the cluster with the highest cluster membership probability. The individual area under the curve (AUC) for the OGTT was adopted to evaluate the severity of maternal hyperglycemia by summing the area of three trapezoids as follows: (0-h + 1-h glucose)/2, (1-h + 2-h glucose)/2, and (2-h + 3-h glucose)/2. Binary logistic regression analysis was performed to compare the prevalence of outcomes between the normal and three latent glucose pattern groups, four groups by classifying quartiles of individual AUCs, or three groups according to the number of criteria in GDM patients. Two multiple logistic regression models were used to control for the confounding factors. Model A included age, preexisting hypertension, family history of diabetes mellitus, family history of hypertension, pre-pregnancy body mass index (BMI), parity, and gestational age before delivery as covariates. Model B additionally included SBP, glucose level at 35 weeks, and insulin treatment. The risk associated with the outcome was calculated and presented as the OR and corresponding 95% CI. We also used a restricted cubic spline (RCS) curve with four knots for the adjusted ORs to graphically demonstrate the nonlinear relationship between the individual AUC for OGTT and the risk of adverse pregnancy outcomes. All reported p -values were two-sided, and statistical significance was set at p <0.05. We used the Mclust function in the mclust package (version 5.4.6) in R software (version 4.0.3; R Foundation for Statistical Computing, Vienna, Austria) to conduct the LPA. 21 All other statistical analyses were performed using SAS software (version 9.4; SAS Institute Inc., Cary, NC, USA).