Research On Risk Estimationn In Down Syndrome Prenatal Screening At Second Trimester

Objective The aim of this study was to establish appropriate biostatistics methods at second trimester prenatal screening for Down Syndrome, including correction model for maternal serum markers by gestational age, weight correction model, risk estimation model.The models were based on the data of Chinese population at second trimester. The effect of prenatal screening was estimated by different approaches of gestational age evaluation.Methods Weighted Polynomial Regression models were utilized to regress maternal serum markers and gestational age. The optimal models were selected by PRESS Statistics and R2adj. Then the models were compared with those built in screening software Lifecycle 3.0 and 2 T-risk by cumulative sum plots. Weighted Polynomial Regression models were used to regress MoM of maternal serum markers with weight. The risk of having affected fetus was calculated in Logistic regression and gaussian distribution method, including triple marker test model and double marker test model. The effects of Logistic model and guassion model were compared by ROC curve analysis with 2T-risk. The sensitivity, specificity and the area under ROC curve were compared.Results The optimal models between maternal serum markers and gestational age were well fitted. The cumulative sum of log10(MoM of AFP), log10(MoM of Freeβ-HCG) and log10(MoM of uE3) for the models was near to zero while those values for Lifecycle 3.0 and 2 T were far deviated to zero. The median of MoM of AFP,Freeβ-hCG and uE3 corrected by new models were approximate 1, better than 2T-risk. The means for MoM of AFP,Freeβ-hCG and uE3 were 1.000513, 1.002751 1.002599, otherwise the ones for 2T-risk's were 0.925778, 0.869931, 1.059153. There were statistically significant, revealted by paired t test of each index. The effect of Logistic model and guassion model were better than 2T-risk, using the sensitivity, specificity and the area under ROC curve. At 5% false positive rate ,the detection rate of guassion risk estimation model was 78.1%, the cutoff value was 330. The detection rate of Logistic risk estimation moedl was 75%, the cut-off value was 450. The detection rate of 2T-risk was 71.9%, the cut-off value was 550.Conclusion Our gestational age correction model for maternal serum markers, weight correction model and risk estimation were more applicable than the models built in 2T-risk at second trimester screening.