Cutoff points of TG at each gestational week
ROC curves for the prediction of gestational TG level on hypertriglyceridemia at 42 days postpartum reached an AUC of 0.759, 0.750, 0.738 and 0.708 at 6-8th,16th, 24thand 36th gestational week, respectively (Fig. 2 ). TG cutoff point was 1.12 mmol/L at 6-8thgestational week (sensitivity = 67.7%, specificity = 70.7%), 1.93 mmol/L at 16th gestational week (sensitivity = 61.7%, specificity = 75.9%), 2.35 mmol/L at 24th gestational week (sensitivity = 66.2%, specificity = 73.4%), and 3.08 mmol/L at 36th gestational week (sensitivity = 70.1%, specificity = 63.8%) (Table 6) . According to AUC and the Hosmer-Lemeshow goodness-of-fit test, the prediction model had favorable discrimination and calibration (χ2 = 6.156, P = 0.630; χ2 = 10.382, P = 0.239; χ2 = 9.786, P = 0.280; χ2 = 21.917, P = 0.050) (Table 6) .
Effects of pre-BMI and GDM on the risk ofpostpartum hypertriglyceridemia
According to the hierarchical logistic regression analysis for different pre-BMI subgroup, the positive association between gestational TG levels and risk of postpartum hypertriglyceridemia remained in normal weight and overweight & obese group. At 16th, 24th and 36th gestational week, the risk of postpartum hypertriglyceridemia for women with normal pre-BMI [OR 3.756, 95% CI (2.337-6.035); OR 2.603, 95% CI (1.760-3.850); OR 2.174, 95% CI (1.630-2.901)] was higher than overweight & obese group [OR 2.748, 95% CI (1.457-5.181); OR 2.451, 95% CI (1.505-3.991); OR 1.862, 95% CI (1.295-2.677)], respectively(Table 7) .
As for the subgroup divided by GDM, the risk of postpartum hypertriglyceridemia for non-GDM women at 6-8th, 16th, 24th and 36th gestational week [OR 5.710, 95% CI (2.975-10.956); OR 3.421, 95% CI (2.208-5.302); OR 2.580, 95% CI (1.833-3.632); OR 2.161, 95% CI (1.646-2.836)] was higher than GDM group [OR 3.702, 95% CI (1.673-8.193); OR 2.890, 95% CI (1.465-5.701); OR 2.125, 95% CI (1.216-3.714); OR 1.821, 95% CI (1.253-2.646)], respectively (Table 7) .