Establishment and valuation of predictive model for gestational diabetes mellitus

Objective To establish risk assessment model for gestational diabetes mellitus (GDM) based on conventional clinical index,in order to effectively prevent and control GDM.Methods This prospective study was conducted in 1765 pregnant woman who were registered in Yuzhou People's Hospital from December 2016 to August 2018. The age,pre-pregnancy body mass index (BMI),gestational weight gain,parity,bad history of pregnancy and childbirth,family history of diabetes,plasma lipids,fasting plasma glucose (FPG) of the participants during the first trimester were collected. According to whether or not GDM,the pregnant women were divided into GDM group (n = 157) and non-GDM group (n = 1608). The above-mentioned index were compared between two groups with univariate and multivariate analyses. A risk prediction model of GDM was established.Results ① Univariate analysis indicated that the age,pre-pregnancy BMI,gestational weight gain,parity,family history of diabetes,triglyceride (TG),FPG during the first trimester had statistical difference between GDM group and non-GDM group. ② Multiple logistic regression analysis showed that age,prepregnancy BMI,gestational weight gain,TG,FPG during the first trimester were the independent prediction index of GDM (P<0. 05).Prediction model was as follows: P_GDM= 1/ 1 + EXP-(-8. 892 + 0. 203 × gae (25-34 years) + 1. 085 × age≥35 years-0. 810 ×pre-pregnancy lean + 0. 992 × pre-pregnancy over weight + 1. 938 × pre-pregnancy obesity-0. 740 × insufficient gestational weight gain +1. 169 × excessive gestational weight gain + 0. 643 × TG + 0. 906 × FPG) . ③ The area under the receiver operating characteristic (ROC) curve of mode which were used to predict GDM was 0. 824 (95% CI: 0. 793-0. 856). Compared with the random area (0. 5),there was statistical difference (P= 0. 000). The cut off point for prediction probability was 0. 532 (Youden's index was the biggest),and the sensibility,specificity and accuracy of mode for GDM was 0. 733,0. 796 and 79. 04%,respectively. When the cut off point was 0. 5,the sensibility,specificity and accuracy was 0. 814,0. 656 and 67. 08%,respectively.Conclusion The risk prediction model based on the factors such as age,pre-pregnancy BMI,gestational weight gain,TG,and FPG during the first trimester can provide reference for early warning of GDM.