Fig. 2

a Levels of FPG and TG by gestational age between 8th and 20th week Number of participants at each gestational week indicated at the top of the figure. Figure 2b ROC curve of GDM risk prediction models based on multivariate logistic regression. The black curve was derived from regression model using categorical variables as indicated in Table 3 (rightmost column). The curve in blue was derived from model following Bayesian variable selection and model averaging and, this best fitted model using maternal age, pre-gravid BMI, FBG, TG, TC and LDL as predictive variables. The predicted optimal cut-off probabilities to discriminate GDM from non-GDM mothers for categorical variable model and Bayesian variable selected model were 0.083 (with specificity = 0.80 and sensitivity = 0.58) and 0.118 (with specificity = 0.75 and sensitivity = 0.66), respectively. The AUR was presented as averaged level and 95% CI in the same color