Objective To evaluate the differences of α-fetoprotein (AFP), ffeeβ-human chorionic gnnadotropin (HCG) indexes in 3 foreign median databases for antepartum risk screening, and establish the median databases of normal pregnant women in Wenzhou for antepartum screening of AFP, free-β-HCG indexes through the suitable median computational models. Methods The levels of AFP and free β-HCG of 20054 normal pregnant women in Wenzhou were detected by time-resolved fluorometry. The data in this paper were compared with the data of 2T-risk ( 2T), Lifecycle-2. 2 (LC2.2 ) and Lifecycle -3.0 (LC3.0) by double-factor ANOVA. The differences between the data in the paper and the data from Shenyang and Ningbo were analyzed. The median database of Wenzhou pregnant women was established by the suitable regression model, with the stability of nonlinear regression models of the 3 software assessed by model correcting fitting, distribution mean of model fitting logarithmic and standard deviation. Results The levels of AFP and freeβ- hCG reported here were 10% and 16% higher than the data of 2T-risk, 15% and 20% higher than that of LC 2. 2, 6% and 17% higher than that of LC 3.0 respectively. The differences were statistically significant. ( FAFP = 161. 757 ,P < 0. 01 ; Ffree-β-HCG = 58. 261, P < 0.01 ). The levels of AFP and free β- hCG in Wenzhou were 2% higher and 3% lower than that of Shenyang, 1% and 2% higher than that of Ningbo. There was no statistical difference of AFP levels among Wenzhou , Shenyang and Ningbo( FAFP = 0. 174 ,P =0. 840) while the differences of free-β-hCG were statistically significant( F<,free-β-HCG> = 13.303 ,P < 0. 01 ). The differences of quadratic equation regression model, exponent quadratic function regression model and exponent quadru-function regression model of 2T, LC-2. 2 and LC-3.0 were not remarkable. The exponent quadru-function regression model was the best. Conclusions There are significant differences between the data from Wenzhou, Shenyang and Ningbo and the data of T-2 risk, LC-2. 2, LC -3.0. The discrepancy is due to the ethnic and different parameters of regression models. So the model parameters and the median databases are urgently required in China. The differences of large sample size of AFP from Wenzhow, Shengyang and Ningbo are not significant, while the differences of free-β- hCG from Wenzhow, Shengyang and Ningbo is remained because of its instability. The levels from Wenzhow and Ningbo are near. It is suggested that the laboratories with small sample size can establish their own laboratory parameters using the reference obtained from large sample size under the same experimental conditions. There are no significant differences of stability among regression computational models in the 3 software. The exponent quadru-function regression model can be used to establish the median databases for the screening with the similar data distribution in the paper.