Establishment of gestational diabetes risk prediction model and clinical verification

Abstract

ObjectiveThe present study aimed to evaluate the risk factors for gestational diabetes mellitus (GDM) and build and validate an early risk prediction model of GDM by comparing the differences in the indicators of the first trimester of pregnancy between pregnant women with GDM and non-gestational diabetes mellitus (NGDM). Thus, this study provided a theoretical basis for early intervention of GDM.MethodsA total of 6000 pregnant women who underwent a routine prenatal examination in Qinhuangdao Maternal and Child Health Hospital (Qinhuangdao City, Hebei Province, China) from January 2016–2022 were retrospectively selected and randomly divided into a modeling cohort (4200 cases) and validation cohort (1800 cases) at a ratio of 3:7. According to the results of oral glucose tolerance test (OGTT), they were divided into NGDM and GDM groups. The modeling cohort consisted of 2975 NGDM and 1225 GDM cases, while the validation cohort consisted of 1281 NGDM and 519 GDM cases. The differences in general conditions and laboratory indicators between different groups were compared, and logistic regression analysis was further used to establish a risk prediction model for GDM in the first trimester. The receiver operating characteristic curve (ROC) and Hosmer–Lemeshow (HL) tests were used to evaluate the prediction of the model efficacy.ResultsAge, pre-pregnancy body mass index (BMI), glycosylated hemoglobin (HbA1c), blood uric acid (UA), triglyceride (TG), and high-density lipoprotein cholesterol (HDL-C) in the first trimester were independent risk factors for GDM (P < 0.05). The model equation was Y = 1/{1 + exp[− (− 18.373 + age × 0.065 + BMI × 0.030 + first-trimester HbA1c × 2.519 + UA × 0.014 + TG × 0.224-HDL-C × 0.635)]}. The area under the ROC curve (AUC) of the model cohort was 0.803 (0.788–0.817), the sensitivity was 72.0%, and the specificity was 73.5%. The AUC of the validation cohort was 0.782 (0.759–0.806), the sensitivity was 68.6%, and the specificity was 73.8%. The P values of the HL test in both the training and validation sets were > 0.05, indicating a satisfactory model fit.ConclusionAge, pre-pregnancy BMI, HbA1C in early pregnancy, blood UA, TG, and HDL-C are independent risk factors for GDM. The risk prediction model established by combining age, pre-pregnancy BMI, and laboratory indicators in the first trimester can provide a theoretical basis for early screening, monitoring, and intervention of GDM high-risk pregnant women.